Datasets:

Hantao

/

ChemImages

like 1

4147_cplu201900281-fig-0044.jpg

Figure 44Open in figure viewerPowerPointCore (cellulose) – sheath (PVDF-HFP) nanofiber separators: (a) preparation process from cigarette filter to nanofibers; (b) TEM observation of prepared core-sheath fibers; (c) as-prepared separators (Celgard 2300 commercial separator, cellulose fiber separator, PVDF-HFP fiber separator, and cellulose core & PVDF-HFP sheath fiber separator); (d) corresponding separator conditions after heat treatment in at 200 °C for 1 hr. Reprinted with permission from Ref [172]. Copyright 2015 American Chemical Society.

An interesting eco-friendly cellulose-based membrane separator for high-performance lithium-ion batteries was developed using waste cigarette filters. Cellulose has some attractive properties as a LIB separator, including thermal stability and hydrophilicity. Due to its excellent thermal stability, cellulose acetate is being used for cigarette filters. Recycled cigarette filter was dissolved in core solution, while PVDF-HFP polymer solution used as a sheath, as shown in Fig 44a. Cellulose core and PVDF-HFP sheath structure was very well formed as observed by TEM (Figure #cplu201900281-fig-0044#44b) and exhibited good tensile strength ∼34.1 MPa, high porosity ∼66 %, excellent thermal stability, and electrolyte compatibility. The thermal stability is demonstrated by observing before (Figure #cplu201900281-fig-0044#44c) and after (Figure #cplu201900281-fig-0044#44d) heat treatment. Commercial separator and PVDF-HFP fiber are severely shrunk while the core-sheath separator remains intact. Compared to commercial separator, it also shows lower interfacial resistance ∼98.5 Ω, higher ionic conductivity ∼6.16 mS/m, superior rate capability ∼138 mA-h/g and cycling performance ∼75.4 % after 100 cycles.

4147_cplu201900281-fig-0045.jpg

Figure 45Open in figure viewerPowerPointMorphology observations of core-sheath TiN-VN fibers using (a) SEM and (b) TEM. (c) Specific capacitances of electrodes with TiN, VN, and TiN-VN fibers for scan rates from 2 to 50 mV/s. Reprinted with permission from Ref [174]. Copyright 2011 American Chemical Society. Porous carbon nanofibers: (d) specific capacitance comparison based on porous structures. Reprinted with permission from Ref [175]. Copyright 2015 John Wiley and Sons. BaTiO3 core and TiO2 sheath fiber embedded PVDF film capacitor: (e) Illustration of the composite PVDF film with BaTiO3-TiO2 core-sheath nanofibers showing lattice structures in different parts of the nanofiber. Inset: false color SEM image of the fractured BaTiO3 core-TiO2 sheath fiber. Reprinted with permission from Ref [176]. Copyright 2016 Royal Society of Chemistry.

Compared to rechargeable batteries, supercapacitors have certain advantages, such as simple construction, environmental safety, fast rate capability, high power density, safety and long cycle life. Same benefits from (coaxially) electrospun nanofibers that applied to battery operation also apply to supercapacitor functions. Supercapacitors made of coaxial fibers with titanium nitride (TiN) core and vanadium nitride (VN) sheath have been demonstrated. The core-sheath structure is shown in Figure #cplu201900281-fig-0045#45a and b. The fibers have a very porous structure that is beneficial for increased capacity. The TiN core exhibits a better electronic conductivity with low capacity, while the vanadium nitride (VN) sheath possesses a higher capacity with poor electronic conductivity. Combining better rate capability of TiN and higher specific capacitance of VN into core-sheath fibers, the coaxial fiber can provide a specific capacitance of 247.5 F/g at the scan rate of 2 mV/s and a better rate capability of 160.8 F/g at the higher scan rate of 50 mV/s, as shown in Figure #cplu201900281-fig-0045#45c. Porous hollow carbon nanofibers made by carbonizing SAN core and polyacrylic acid (PAA)/PVP sheath also provided very high specific capacitance of 221 F/g with superior capacitance retention of 95 % after 5000 cycles at a scan rate of 0.1 V/s. PVP in the sheath was used as a pore inducer. Hollow core and porous sheath enhanced the capacitance ∼20 % and 47 %, respectively (Figure #cplu201900281-fig-0045#45d).

4147_cplu201900281-fig-0046.jpg

Figure 46Open in figure viewerPowerPointWS2 single layer anchored hollow carbon nanofibers: (a) illustration of prepared fiber and (b) TEM image of WS2 single layer anchored on carbon wall; (c) Tafel plots for the WS2-based electrocatalysts and a commercial Pt/C electrocatalyst. Reprinted with permission from Ref [178]. Copyright 2015 American Chemical Society.

Hydrogen is regarded as the ideal clean energy source due to the largest energy density and absence of CO2 emission. Most hydrogen in use is produced by fossil fuels (typically natural gas – CH4) in a process that releases a significant level of carbon dioxide byproduct. An alternative promising method is the environmentally friendly electrochemical water splitting. The hydrogen evolution reaction (HER, 2 H++2 e−→H2) is the cathodic half reaction of electrochemical water splitting. For high energetic efficiency in water splitting, an electrocatalyst is required to minimize the overpotential and initiate the electrochemical reaction. Pt is the most efficient electrocatalyst but is hardly cost-effective. Hollow nitrogen-doped carbon nanofibers anchored with single-layered tungsten disulfide (WS2) nanoplates was developed as a low cost and efficient electrocatalyst for enhanced HER. In coaxial electrospinning, SAN core provided robust support for the sheath structure due to immiscibility with WS2 precursor dispersed PAN sheath solutions. In two step thermal treatments, single layer WS2 nanoplates were formed at 400 °C and hollow carbon fibers were finalized at 700 °C (illustrated in Figure #cplu201900281-fig-0046#46a). Nanopores from 2 to 50 nm in diameter were also formed on hollow carbon wall by generated sulfur gases and decomposed SAN core vapor during thermal treatments. Formation of WS2 single layer nanoplates ∼5 nm in length were confirmed in TEM observation (Figure #cplu201900281-fig-0046#46b). Compared to the WS2 single layer nanoplate anchored solid carbon nanofibers (WS2@NCNFs) and WS2 powders, WS2 single layer nanoplate anchored hollow carbon nanofibers (WS2@HNCNFs) provides lower overpotential of 280 mV to reach −10 mA/cm2 with lower charge transfer resistance of 24 Ω, a small Tafel slope of 60 mV/dec (Figure #cplu201900281-fig-0046#46c), and excellent durability, which are beneficial for enhancing HER performance.

4147_cplu201900281-fig-0047.jpg

Figure 47Open in figure viewerPowerPointElectronic applications of nanofibers and their characteristics: (a) fiber conductivity (log scale) vs. weight fraction of PAni in electrospun fibers. Reprinted with permission from Ref [185]. Copyright 2012 American Chemical Society. (b) Process to fabricate aligned P3HT nanofiber FET and (c) FET output characteristics. Reprinted with permission from Ref [186]. Copyright 2011 American Chemical Society.

Although dielectric capacitors have very attractive features such as fast charging−discharging speeds and high-power density, achieving high energy density is needed in order to be used for many applications. Lin et al. achieved high energy density in film capacitor by incorporating BaTiO3-TiO2 core-sheath fibers into PVDF film as shown in Figure #cplu201900281-fig-0047#47e. Charge shifting between core and sheath induces additional interfacial polarization and increases electric displacement in the film, while the breakdown strength is maintained because the charge shifting is limited to the local interfacial zone. This nanocomposite film with 3 % volume fraction of core-sheath fibers provided a large energy density ∼10.94 J/cm3 at a field of 360 kV/mm, which is more than 2 and 4 times higher than PVDF and commercial BOPP (biaxially oriented polypropylene) films, respectively. Recently, coaxially electrospun BaTiO3 nanotubes coated with a dense dopamine layer were incorporated into PVDF films to improve dielectric properties and energy density of the polymer composite material. Dopamine coating enables uniform distribution of nanotubes within the polymer film without having any aggregation and also confines the charge carrier movements in the interface between nanotubes and polymer host, resulting in enhanced dielectric properties of the composite film. Composite films (10.8 vol.% BaTiO3 nanotubes) provided 5.7× dielectric constant (47.5) than the pristine PVDF of 8.26. Also, the energy density of 7.03 J/cm3 at relatively lower field of 330 MV/m is 6.25× higher than 1.2 J/cm3 at 640 MV/m of the commercial biaxially oriented polypropylene (BOPP). Most conductive nanofibers consist of non-polymeric materials obtained through a high temperature process. Conductive polymers such as polyaniline (PANi), polypyrrole (PPy), poly(3,4-ethylenedioxythiophene) (PEDOT) are not electrospinnable due to their low molecular weight and non-elasticity of their polymer chain. Therefore, they have been frequently blended with non-conducting electrospinnable polymers to produce the conductive nanofibers, although their conductivity was dramatically reduced. Coaxial electrospinning can be utilized to form a pure conductive fiber without a host polymer. Zhang and Rutledge demonstrated pure PAni nanofibers formed by coaxial electrospinning after selective removal of PMMA sheath used for fiber formation. As shown in Figure #cplu201900281-fig-0047#47a, pure PAni fiber shows excellent fiber conductivity of 50±30 S/cm. Interestingly, added polymers also affected the blended fiber conductivity due to the difference of their intrinsic conductivities and PAni compatibility. PEO blended PAni showed higher conductivity and better fiber formation than that of PMMA. Further aligning PAni molecules to the axis of current flow by simple mechanical stretching enhanced the conductivity up to 130±40 S/cm. One-dimensional nanofiber would significantly improve the charge mobility by enabling anisotropic charge transport and better molecular arrangement. P3HT has been popularly used as a semiconducting material for organic field effect transistor (OFET) due to superior carrier mobilities up to 0.1 cm2/(V-s). However, due to the difficulty of electrospinning P3HT, conventional single nozzle electrospinning could not produce nanofibers with consistent diameter without bead formation. Lee et al. successfully demonstrated continuous P3HT fibers using coaxial electrospinning. A P3HT chloroform solution was used for the core and chloroform solvent was also used for the sheath to prevent early P3HT precipitation at the nozzle. Field effect transistors fabricated with pure P3HT fibers resulted in good charge mobility of 0.017 cm2/(V-s), which is ∼10× higher than P3HT : PCL (8 : 2 wt. ratio) blended fibers of ∼0.0012 cm2/(V-s). This approach would be also useful for other materials in highly volatile solvents. Chen et al. also demonstrated organic FET (OFET) using core-sheath fibers with P3HT core and PMMA sacrificial sheath. As shown in Figure #cplu201900281-fig-0047#47b, core-sheath fibers were uniaxially deposited and then transferred to a 200 nm-thick SiO2 coated heavily n-doped silicon substrate. After PMMA sheath removal using selective solvent and thermal annealing, 350 nm-thick source and drain gold electrodes were deposited on top of fibers, providing a good contact between fibers and electrodes. Lower flow rate of PMMA sheath solution induced enhanced π-π stacking and crystallinity of P3HT, which dramatically improved the charge mobility of fibers to 0.19 cm cm2/(V-s). Optimizing thermal annealing temperature is also important to prevent the relaxation of P3HT molecular orientation.

4147_cplu201900281-fig-0048.jpg

Figure 48Open in figure viewerPowerPointLight emitting coaxial fibers: (a) structural and operational schematic; (b) luminescence response based on device applied voltage in N2 environment. Reprinted with permission from Ref [190]. Copyright 2012 American Chemical Society.

Light-emitting nanofibers can be useful for integration of optoelectronic devices into smart textiles. Previously, single nozzle electrospun light-emitting fibers embedded with ruthenium-based ionic transition-metal complex (iTMC) were demonstrated. Electrospun fibers deposited between two electrodes produced luminescence upon applying bias. Using coaxial electrospinning, Yang et al. successfully integrated two electrodes into coaxial structure, as illustrated in Figure #cplu201900281-fig-0048#48a. Highly conductive Galinstan (containing Ga, In and Sn) liquid metal was used for the core (cathode) and the mixture of ruthenium(II) tris(bipyridine) and PEO was used for the electroluminescent sheath. A transparent and conducting ITO thin layer (anode) was evaporated on the core-sheath fibers. While single electroluminescent fibers only produced light within the narrow gap between two electrodes, core-sheath fibers can produce electroluminescence over the entire fiber length. Higher applied bias can produce stronger luminescence, as demonstrated in Figure #cplu201900281-fig-0048#48b. These self-supporting fibers provide one-dimensional flexible, lightweight, and conformable light sources that have potential applications in optoelectronic textiles, bioimaging, flexible display, etc.

4147_cplu201900281-fig-0049.jpg

Figure 49Open in figure viewerPowerPointTriaxial electrospinning: (a) experimental configuration. Reprinted with permission from Ref [107]. Copyright 2013 American Chemical Society. Fiber-in-tube structures: (b) SEM photo; (c) TEM photo. Reprinted with permission from Ref [194]. Copyright 2010 American Chemical Society. Triaxially electrosprayed microcapsules: (d) SEM morphology; (e) laser confocal image. Reprinted with permission from Ref [195]. Copyright 2011 Elsevier.

Tri-layered coaxial (core-intermediate-sheath) fibers produced by electrospinning using a spinneret with three coaxially aligned inlets (aka “triaxial electrospinning”) have emerged recently due to their novel and unique benefits: (a) combination of three different sets of material properties; (b) enhanced separation between core and sheath layers; (c) stronger binding between two materials from the sandwich configuration; (d) controlled release of encapsulated materials regardless of sheath surface properties. The basic fiber formation mechanisms are the same as with coaxial electrospinning, although additional parameters related with interactions between solutions need to be considered. A third syringe pump set is required to feed the triaxial nozzle, as illustrated in Figure #cplu201900281-fig-0049#49a. Although this variation of fiber electrospinning is a nascent field, the flexibility and versatility that triaxial electrospinning offers makes it very likely that it will develop rapidly in the near future. Selected examples of fiber formation and related applications using triaxial electrospinning are summarized in Table 4. Chen et al. successfully demonstrated the unique nanowire-in-tube structure shown in Figure #cplu201900281-fig-0049#49b and c. TiO2 precursor was used for core and sheath, while a paraffin emulsion solution was used for the sacrificial intermediate layer. Electrosprayed tri-layered microparticles were demonstrated by Kim and Kim. For triaxial electrospinning, PLGA solution was fed into both core and sheath nozzle outlets, while PDLLA solution was fed into the intermediate outlet. Very uniform monodisperse-sized microcapsules were obtained (Figure #cplu201900281-fig-0049#49d) and their concentric tri-layered structure was observed using laser confocal microscopy (Figure #cplu201900281-fig-0049#49e). The capsule size and shell thickness can be easily manipulated by adjusting flow rates and polymer concentrations. A slightly modified version of multiple input electrospinning was reported by Zhao et al., as shown in Figure #cplu201900281-fig-0050#50a. Fibers with multiple parallel hollow channels were formed using multiple inner fluids fed in parallel rather than coaxially. This interesting internal structure is displayed in the SEM photographs of Figure #cplu201900281-fig-0050#50b. By varying the number of input solutions, hollow fibers with 2, 3, 4, and 5 parallel hollow channels were demonstrated. These unique structures including fiber-in-tube structure (Figure #cplu201900281-fig-0049#49b) are potentially very useful for various applications, such as energy, electronics, thermal barriers, etc.

4147_cplu201900281-fig-0050.jpg

Figure 50Open in figure viewerPowerPointMulti-channel hollow microtubes by multiple input electrospinning: (a) experimental configuration; (b) cross-sectional SEM photographs of fibers with different numbers of inner fluid channels. Reprinted with permission from Ref [201]. Copyright 2007 American Chemical Society.

A slightly modified version of multiple input electrospinning was reported by Zhao et al., as shown in Figure #cplu201900281-fig-0050#50a. Fibers with multiple parallel hollow channels were formed using multiple inner fluids fed in parallel rather than coaxially. This interesting internal structure is displayed in the SEM photographs of Figure #cplu201900281-fig-0050#50b. By varying the number of input solutions, hollow fibers with 2, 3, 4, and 5 parallel hollow channels were demonstrated. These unique structures including fiber-in-tube structure (Figure #cplu201900281-fig-0049#49b) are potentially very useful for various applications, such as energy, electronics, thermal barriers, etc.

4147_cplu201900281-fig-0051.jpg

Figure 51Open in figure viewerPowerPointFree surface electrospinning using wire electrodes: (a) wire electrode/bath configuration; (b) TEM photograph of resultant fibers with PEO/PEG core and PS sheath; (c) evolution of surface profiles of two immiscible liquids as wire (viewed end-on) travels through liquid interfaces. Reprinted with permission from Ref [202]. Copyright 2013 Elsevier. Slit-surface electrospinning: (d) slit configuration (top view); (e) slit unit; (f) compound coaxial jet ejection (side view); (g) resultant core-sheath fiber; (h) resultant hollow fiber. Reprinted with permission from Ref [204]. Copyright 2015 Public Library of Science.

As mentioned earlier in Introduction (Figure #cplu201900281-fig-0009#9), mass production of single electrospun nanofibers has been well developed in last decade. However, mass production of core-sheath fibers via coaxial electrospinning is more challenging because of the complexity of dealing with two different solutions in separate layers. In addition to clogging issues and the repulsive interaction between adjacent liquid jets, sufficiently high production rate for commercial needs cannot be achieved by integrating several nozzles. Nozzleless electrospinning, which can generate hundreds to thousands liquid jets without mechanical complications is more suitable for large-scale mass production of fiber mats than conventional nozzle-based electrospinning. Innovative techniques are being explored for adapting nozzleless electrospinning to the formation of core-sheath fibers. Forward et al. developed a nozzleless coaxial electrospinning approach using wire electrodes (Figure #cplu201900281-fig-0051#51a). As illustrated in Figure #cplu201900281-fig-0051#51c, a wire electrode is drawn through a liquid bath with two layered immiscible liquids. As the wire sweeps through the bath, it is first coated with the bottom liquid and then with the top liquid, forming a compound liquid layer which then breaks into compound droplets on the wire, producing core-sheath fibers (Figure #cplu201900281-fig-0051#51b). Another approach for nozzleless coaxial electrospinning was proposed using a weir spinneret and a slit spinneret (Figure #cplu201900281-fig-0051#51d–e). For the slit coaxial electrospinning, solutions are fed into the core and sheath slits separately and form multiple compound coaxial liquid jets, as shown in Figure #cplu201900281-fig-0051#51f. Core-sheath fibers (Figure #cplu201900281-fig-0051#51g) and hollow fibers (Figure #cplu201900281-fig-0051#51h) were successfully obtained using this method. However, although the potential for scale up of electrospun core/shell fiber production is greatly enhanced, there currently still exist some limitations, such as solvent immiscibility, vapor pressure differences, formation of defect-free sheath layers, that need to be solved. Recently, air-blowing assisted coaxial electrospinning using a triaxial nozzle was demonstrated to provide higher production rate of core-sheath/hollow nanofibers by Duan and Greiner. Air-blowing enables much higher flow rate of fed solutions, providing the total flow rate of 5.8–12 mL/hr and the productivity up to 3.6 g/hr, which is much higher than conventional coaxial electrospinning.

4147_cplu201900281-fig-0052.jpg

Figure 52Open in figure viewerPowerPointNear-field coaxial electrospinning: (a) hollow PCL fibers (core: sugar); (b) patterned grid of sugar-PCL core-sheath fibers. Reprinted with permission from Ref [209]. Copyright 2011 Elsevier.; (c) illustration of near-field coaxial electrospinning on a rotating collector. Resultant hollow PVDF fibers produced with air flow rate of (d) 12 mL/hr; (e) 15 mL/hr. Reprinted with permission from Ref [210]. Copyright 2015 RSC Publishing.

A comprehensive review of near-field electrospinning was recently reported by the Long group. Near-field electrospinning was also applied to coaxial electrospinning and produced aligned and patterned sugar-PCL core-sheath fibers, as shown in Figure #cplu201900281-fig-0052#52a and b. While the motorized x-y stage has been mostly used for moving the collector, Pan et al. utilized the rotating collector fixed on x-y stage as shown in Figure #cplu201900281-fig-0052#52c, which can provide relatively fast lateral speed and conveniently collect many aligned fibers. The gap between aligned fibers is controlled by an X−Y stage under the rotating collector. With air being pumped through the core of the nozzle, hollow piezoelectric PVDF fibers were formed (Figure #cplu201900281-fig-0052#52d and e), which can be challenging in conventional coaxial electrospinning. Hollow piezoelectric PVDF fibers provided ∼2.5× times higher power generation than the solid piezoelectric PVDF fibers.

4147_cplu201900281-fig-0053.jpg

Figure 53Open in figure viewerPowerPointApplication tree for fiber membranes formed by electrospinning.

The characteristic features of coaxial electrospinning in the formation of fibers with unique and versatile properties, such as multifunctional materials properties, in/organic and biological material encapsulation, complex multilayered (solid or hollow) structure and geometry, has enabled the field to grow at a rapidly increasing pace, expanding and bearing fruit in various areas, as illustrated in Figure #cplu201900281-fig-0053#53. Moreover, the recent addition of mass production capability is accelerating the commercialization of novel core-sheath nanofiber products.

16368_chem202001312-fig-5001.jpg

Scheme 1Open in figure viewerPowerPointProposed synthetic building block to obtained novel fluorescent (RGD(W/w)(7-3-Ptz)) derivatives through intramolecular C−C Suzuki–Miyaura cross-coupling.

Although it is a difficult task which depends on conformational stability, the atropisomers could be identified by NMR. In this work the P-isomers could be successfully isolated as major atropisomers in the case of 12 c–f and characterized by NOESY experiments. The chemical shifts below 5 ppm of the α-proton of tryptophan in the rigid cyclopeptides 12 c–f suggest a P-biaryl isomer (Figure #chem202001312-fig-0001#1 and Supporting Information). NOE proximities were found between the NH1 and H2 of indole and both the α-proton of aspartic acid and the amide proton of tryptophan, suggesting a more constrained cycle with the indole ring facing inside. It seems that the rigidity of the structure in the smaller cyclopeptides 12 c,d and those with N-methylated tryptophan 12 e,f confers the observed conformational stability. On the other hand, for compounds 12 a,b containing a longer and more flexible linker, separate signals for the atropisomers were not observed.

16368_chem202001312-fig-5002.jpg

Scheme 2Open in figure viewerPowerPointReagents and conditions for the synthesis of Ptz1, Ptz2 and Ptz3: (i) 6-bromohexanenitrile, KOH, NaI, DMF, 50 °C, 48 h, 71 %; (ii) Ethyl bromoacetate, KOH, NaI, DMF, 50 °C, 48 h, 46 %; (iii) MeOH/EtOH/KOH (4 m), reflux, 72 h, 72 %; (iv) & (v) B2Pin2 (2.2 equiv), PdCl2(PPh3)2 (0.02 equiv), KOAc (5 equiv), dry toluene (40 mL), 110 °C, Ar, 18 h, 70 %; (vi) MeOH/EtOH/KOH (eq) (4 m), 60 °C, 30 min, 33 %; (vii) DMF (1.1 equiv), 1,2-DCE, POCl3 (1.2 equiv), 90 °C, 48 h, 46 %;42 (viii) MeOH/EtOH/KOH (4 m), reflux, 72 h, 29 %; (ix) B2Pin2 (2.2 equiv), PdCl2(PPh3)2 (0.02 equiv), KOAc (5 equiv), dry toluene (40 mL), 110 °C, Ar, 18 h, 93 %.

To rationalize the difference in binding affinity of the new derivatives in comparison to Cilengitide, and the effect of the oxidized derivative 12 h in comparison to its precursor 12 c, we performed docking and modeling studies. Docking of Cilengitide to αVβ3 was performed and the resulting complex was similar to the Cilengitide-αVβ3 X-ray single-crystal structure (Figure #chem202001312-fig-0002#2 A, PDB: 1l5g, Figure S51A). When Cilengitide binds to αVβ3 (Figure #chem202001312-fig-0002#2 A), binding occurs at the interface between subunits αV and β3. While the Arg and Asp side chains point into opposite directions making contacts with αV and β3, resp., the glycine residue lies at the interface between both subunits. In the Cilengitide-αVβ3 complex, the Arg side chain is located in a narrow groove at the top of the propeller domain of the αV chain, where the guanidinium group is being held in place by a bidentate salt bridge to Asp218 and Asp150. On the other hand, the Asp group is completely buried in the complex. Docking of P-atropisomers 12 h (oxidized) and 12 c (non-oxidized) revealed that both compounds bind at the same site as Cilengitide (Figure #chem202001312-fig-0002#2, and Figure S51), but the poses and the contacts with the receptor are different, which might explain the decreased affinity. The Arg moiety in R-12 h is rotated towards the β-subunit having an interaction with Tyr166 and Arg216 while the oxygen of the sulfoxide group interacts with Arg214. Trp is oriented towards the α-subunit and interacts with Asp218 and Ala218. In the case of the non-oxidized 12 c, only two interactions were detected; one between the Arg moiety and Tyr178 of the α-subunit, while the second between the Asp and Arg216 in the β-subunit (Figure #chem202001312-fig-0002#2 C).

16368_chem202001312-fig-5003.jpg

Scheme 3Open in figure viewerPowerPointPeptide cyclization through Suzuki–Miyaura cross-coupling.

Finally, the spectral characteristics of 12 h, its precursor 12 c, and the derivative 12 g were evaluated in DMSO because of the low solubility of synthesized compounds in water. (Figure #chem202001312-fig-0003#3). As expected, the absorption of Ptz-RGD derivatives strongly depends on the substitution pattern. Compound 12 h with the oxidized Ptz (sulfoxide) gives rise to two intense absorptions at 286 and 308 nm while the UV spectrum of 12 c has a maximum at 262 nm and a second, weaker absorption band at 308 nm. Compound 12 g, which has an extended π-system in position 10 of the Ptz shows two absorption bands at 268 and 318 nm, together with a new absorption band at 486 nm (Figure #chem202001312-fig-0003#3 A). The three compounds displayed interesting fluorescent properties (Figure #chem202001312-fig-0003#3 B). Importantly, 12 g emits in the red region (693 nm) upon excitation at 489 nm, while excitation of 12 c and 12 h at around 310 nm leads to fluorescence emission at 424 and 454 nm, respectively (Figure #chem202001312-fig-0003#3 B). These experiments demonstrate that RGD peptides containing the Ptz-Trp moiety are suitable compounds to be used in spectroscopic experiments in the visible region and, more importantly, show promise for applications in in vivo imaging in the case of compound 12 g.

16368_chem202001312-fig-5004.jpg

Scheme 4Open in figure viewerPowerPointSynthesis of cyclopeptide 12 g with an additional electron-withdrawing moiety by sequential on resin Suzuki/Knoevenagel reactions.

16368_chem202001312-fig-0001.jpg

Figure 1Open in figure viewerPowerPointMajor atropisomers of compounds 12 c, 12 d, according to 2D NMR (ROESY).

16368_chem202001312-fig-5005.jpg

Scheme 5Open in figure viewerPowerPointOxidation of cyclopeptide c(RGDW(7-3Ptz2)) 12 c to give the sulfoxide 12 h.

16368_chem202001312-fig-0002.jpg

Figure 2Open in figure viewerPowerPointModel for the interaction between integrin αvβ3 and Cilengitide (A), R-12 h (B), and 12 c (C) P-atropisomers. The subunits are represented as follows: α- subunit (pink) and β-subunit (light blue). All the inhibitors interact in the RGD-binding pocket at the interface between α-and β-subunits. In the case of Cilengitide, the model of the crystal structure is presented. The corresponding Yasara homology model for Cilengitide interaction, as well as the superimposition between structures, are shown in the Figure S51.

16368_chem202001312-fig-0003.jpg

Figure 3Open in figure viewerPowerPointSpectral evaluation of cyclopeptides 12 c, 12 h, and 12 g in dimethyl sulfoxide. A) Normalized UV/Vis absorption spectra. B) Normalized fluorescence emission intensity spectra. All compounds were measured under the same experimental conditions with varying their excitation wavelength as follows: 12 c (λex=311 nm); 12 h (λex=312 nm), and 12 g (λex=489 nm).

5854_cmdc202100209-fig-0001.jpg

Figure 1Open in figure viewerPowerPointGeneral strategy for the synthesis of hydrazide/hydrazone derivatives.

Dantrolene (DAN; Figure #cmdc202100209-fig-0001#1) is a drug specifically used in the management of malignant hyperthermia, a life-threatening pathology with fatal course. In a recent work, we disclosed new biological activities exerted by DAN, namely inhibition of monoamine oxidase (MAO) B human enzyme with Ki value in the low micromolar range, acetylcholinesterase (AChE), and aggregation of beta amyloid-40 and hexapeptide tau protein sequence PHF6, i. e. two probes of amyloid aggregation in Alzheimer's disease (AD) brain. It is well known the crucial role of MAO isoforms A and B as metabolizing enzymes in modulating the concentration of neurotransmitters, mostly in some severe and chronic neurodegenerative pathologies. This established reputation is strictly related to the substrate and tissue specificity of both isoforms: MAO A selective inhibitors are clinically administered as antidepressants, while MAO B selective inhibition is commonly used for the treatment of the early symptoms of Parkinson's disease. A new outcome of that study was the discovery of the activation by DAN of the carnitine/acylcarnitine carrier (CAC), with EC50 of 9.3 μM for the purified recombinant wild type (WT) protein. This transporter acts through reductive activation and is involved in trafficking of acyl groups into the mitochondria, carried by l-carnitine. Treated with DAN, this transport system facilitates, under oxidative stress (OS) conditions, the restoring of ATP production and thus cell vitality, but also the export of endogenous acetyl-l-carnitine from mitochondria, with consequent neuroprotective effects. The above promising results prompted us to synthesize a number of novel DAN analogues, with the aim of optimizing the inhibitory activity against MAO B and AChE, ultimately improving their pleiotropic pharmacological potential in the treatment of AD and related neurodegenerative syndromes. In this work we investigated in particular: i) the bioisosteric replacement of the NO2 group, which may be toxicophore and precursor for the production of reactive oxygen species (ROS), with CN group; ii) the improvement of the aqueous solubility of DAN by replacing the hydantoin moiety with other (hetero)cyclic moieties bearing protonatable nitrogen(s); iii) the effect of molecular simplification of the three-ring scaffold in DAN, for detecting the minimal pharmacophoric features responsible for the dual activity on MAO A/B and CAC (Figure #cmdc202100209-fig-0001#1). SARs were investigated as thoroughly as possible, even considering the effect of lipophilicity on the enzymes’ inhibition potency. The synthetic pathways chosen for the preparation of the designed compounds are those explored by Snyder and coworkers with slight modifications. The chemical scaffolds were selected to investigate a range of molecular diversity around the structure of DAN, in terms of variation of stereo-electronic and hydrophobic properties. The general strategy of structural modifications is shown in Figure #cmdc202100209-fig-0001#1, whereas the syntheses of compounds are shown in Schemes #cmdc202100209-fig-5001#1–#cmdc202100209-fig-5003#3. Condensations from suitable aldehydes and amino/hydrazide derivatives were performed in DMF/water or acetone/water mixtures, with acidic catalysis and agitation at room temperature. Final compounds were obtained in moderate (20–50 %) to good (70 %) yields, following a simple workup including filtration and purification through either crystallization or column chromatography.

5854_cmdc202100209-fig-0002.jpg

Figure 2Open in figure viewerPowerPointMAO inhibitors structurally related to dantrolene.

Some of the prepared compounds retained the molecular motif of hydantoin, present in DAN and, as thiohydantoin, in the known hypoglycemic drugs rosiglitazone and pioglitazone (Figure #cmdc202100209-fig-0002#2), which also behave as moderate MAO inhibitors. In turn, the molecular pruning gave simple hydrazone and hydrazide derivatives (2-rings series, compounds 14–21) whose structural pattern could be related to that of isocarboxazide (Figure #cmdc202100209-fig-0002#2), an early irreversible MAO inhibitor.

5854_cmdc202100209-fig-5001.jpg

Scheme 1Open in figure viewerPowerPointReagents and conditions: (i) HCl, 0 °C, NaNO2, rt, 30 min; (ii) 2-furaldehyde, CuCl2, acetone, rt; (iii) amine or hydrazine or hydrazide, DMF/water, HCl (cat.), rt, 24 h.

Inhibition kinetics assessed for 9 a competitive mechanism, with Ki equal to 0.50±0.06 μM (Figure #cmdc202100209-fig-0003#3). Taking into account the possible hydrolytic degradation of imine (see stability studies below), inhibition kinetics were determined without preincubation with substrate, as usually done for MAO inhibition experiments.

5854_cmdc202100209-fig-5002.jpg

Scheme 2Open in figure viewerPowerPointReagents and conditions: (i) HCl, 0 °C, NaNO2, rt, 30 min; (ii) 3-methoxybenzaldehyde, CuCl2, acetone, rt; (iii) 2-thiophenecarboxylic acid hydrazide, DMF/water, HCl (cat.), rt, 24 h.

To investigate a possible correlation between MAO inhibitory potency and lipophilicity, we calculated log P with three different programs (ChemDraw 15.0; ALOG PS 2.1; ChemSketch 2017, see Table S1 in Supporting Information) and compared the calculated values with experimental lipophilicity indexes as assessed by reversed-phase (RP) HPLC in isocratic conditions for the majority of compounds. The log of capacity factors (log k’) measured by RP-HPLC using a mixture of methanol/PBS (60 : 40, v/v) as the mobile phase are reported in Tables 1 and 2, along with Clog P values calculated with ChemDraw. These calculated lipophilicity descriptors correlated with the experimental log k’ values (n=13, r2=0.901) better than the other log Ps calculated by ALOG PS and ChemSketch computational tools (r2 equal 0.705 and 0.568, respectively). As shown by the scatter plots in Figure #cmdc202100209-fig-0004#4 and Figure S1 in Supporting Information, there is no evident correlation trend between MAO A/B inhibition data and lipophilicity calculated by the expert system implemented in ChemDraw for compounds achieving finite IC50 values.

5854_cmdc202100209-fig-5003.jpg

Scheme 3Open in figure viewerPowerPointReagents and conditions: (i) hydrazine or hydrazide, acetone/water, HCl (cat.), rt, 24 h.

The immortalized neuroblastoma cells are a widely used model for neuroprotection assays, while dopamine at a concentration of 10 nM acts as an inducer of oxidative stress, being metabolized by MAOs to form hydrogen peroxide. Once hydrolyzed and oxidized, DCF acts as the fluorescing probe of the oxidative stress (OS) state of cells. In the presence of an antioxidant, or even of a MAO inhibitor, ROS burden is lowered and DCF fluorescence decreased. Figure #cmdc202100209-fig-0005#5 shows that 9 was effective in reducing ROS oxidation of DCF. Its activity is superimposable to that of phenelzine, a nonselective MAO inhibitor used as a positive control in this test. The same experiment performed on dantrolene in our previous work gave similar results.

5854_cmdc202100209-fig-0003.jpg

Figure 3Open in figure viewerPowerPointMichaelis-Menten plot (A) and Lineweaver-Burk linearization (B) of inhibition kinetics of hMAO B with compound 9. Image is representative of a single experiment.

The hydrolytic stability of some representative imino/hydrazone/hydrazide derivatives was determined in buffered aqueous media. The stability studies were carried out on compounds 3, 4 and 9, the most active MAO B inhibitors of the 3-rings scaffold series, at a single concentration of 20 μM in 10 mM phosphate buffered saline (PBS) at physiological pH of 7.4, at 37 °C and for 6 h incubation time. The degradation profiles in Figure #cmdc202100209-fig-0006#6A confirmed a substantial stability for 3 and 4, while the imine 9 resulted, at the end of the 6 h of incubation, in a degradation percentage of 76 %. This result is in line with the nature of the functional group of this molecule: in fact, 9 is the only Schiff base within the synthesized series and therefore by itself less stable than the other hydrazone analogues. The hydrolytic degradation of 9 in its starting reagents was confirmed by the comparison of the chromatograms of 9 and 2-chloro-4-hydroxyaniline, obtained in the same chromatographic conditions (Figure #cmdc202100209-fig-0006#6B). Stability studies in buffered solution were performed following an already described procedure, on a Phenomenex C18 column (150×4.6 mm i.d., 3 μm particle size; Phenomenex, Castel Maggiore, Italy) using a mobile phase consisting of a mixture of methanol-water (75 : 25 v/v, with aqueous formic acid 0.1 %). The used flow rate was 0.500 mL/min while the injection volume was 20 μL. Wavelength of UV-Vis detector was adjusted at 254 nm. The chemical stability was evaluated in a phosphate buffer solution pH 7.4 (10 mM HPO42−/H2PO4−; 100 mM NaCl) at 37 °C. Five different concentrations from 0.5 μM to 20 μM were studied in a 2 h time range (data not shown). Each concentration (0.5, 1, 5, 10 and 20 μM) was tested in triplicate starting from 3 different stocks solution (1 mM) prepared separately. The time range was extended to 6 h only for the samples at 20 μM concentration (Figure #cmdc202100209-fig-0006#6a).

5854_cmdc202100209-fig-0004.jpg

Figure 4Open in figure viewerPowerPointPlots of pIC50 values determined toward MAO A (A) and B (B) versus calculated Clog P values (ChemDraw); only compounds with finite IC50 values are shown.

CAC (SLC25 A20) is essential for the transport into mitochondria of acyl moieties as acylcarnitines, where they are processed by β-oxidation pathway. The protein contains six cysteine residues, but two of them, namely C136 and C155, based on the redox state of the protein, are crucial for the regular function of the carrier. In fact, the transporter is active when the two cysteines are in reduced form, while it is inhibited when a C136-C155 disulfide bridge is formed in conditions of OS. One or both cysteines represent also specific targets for various chemical and physiological thiol reducing agents, allowing to modulate the transport activity of the carrier. We demonstrated that DAN led to a significant recovery of the CAC transport activity of the oxidized protein. Herein, we also investigated whether some newly synthesized derivatives, namely the most active MAO B inhibitors 9 and 20, the DAN homologue 14, and the dual MAO B/AChE inhibitor 19, are effective in activating CAC. To calculate the EC50 values from dose-response curves, that is, the concentration which increases the transport activity of the carrier by 50 % compared to the control, a wide concentration range (1–100 μM) was tested (Figure #cmdc202100209-fig-0007#7). The EC50 values measured after 30 min of incubation were 8.2±2.8 μM (comp. 9), 8.4±1.6 μM (14), 8.2±0.57 μM (19), 13±1.8 μM (20), whereas the whole activation of the WT protein was observed at concentrations close to 100 μM for 8, 19 and 20, and 50 μM for 14.

5854_cmdc202100209-fig-0005.jpg

Figure 5Open in figure viewerPowerPointNeuroprotection of SH-SY5Y cells from oxidative insult; DCF-DA assay. Green line, control cells; purple line, 10 nM dopamine; blue line, 10 nM dopamine+10 nM phenelzine; black line, 10 nM dopamine+10 nM 9. The data represent mean±SD of three independent experiments.

To demonstrate that the action of DAN analogues was exerted on the cysteine residues of the CAC, 1 or 50 mM dithioerythritol (DTE), a strong reducing agent, was added to the reconstitution mixture (see Experimental section) in order to mimic the protein at different states of oxidation. The bar plot in Figure #cmdc202100209-fig-0008#8 shows that the tested molecules enabled the protein to recover a significant transport activity, compared to the control, when the protein is more oxidized, i. e., in the presence of 1 mM DTE.

5854_cmdc202100209-fig-0006.jpg

Figure 6Open in figure viewerPowerPoint(A) time-resolved stability of compounds 3, 4 and 9 at the concentration of 20 μM in PBS at 37 °C. Data are representative of three independent experiments and values expressed as mean. (B) overlapping chromatographic peaks of 9 at t 0 and t 4 h and chromatogram of 2-Cl-4-OH aniline.

5854_cmdc202100209-fig-0007.jpg

Figure 7Open in figure viewerPowerPointDose-response curves of activation of CAC (purified recombinant WT protein). The antiport rate was measured by adding 0.1 mM [3H]-carnitine to proteoliposomes containing 15 mM internal carnitine and stopped after 30 min by the specific inhibitor N-ethylmaleimide (NEM). Compounds at increasing concentrations were added 2 min before the transport assay. Values are mean±SD from three independent experiments.

5854_cmdc202100209-fig-0008.jpg

Figure 8Open in figure viewerPowerPointEffects of DAN analogues on the recombinant WT CAC protein. The proteoliposomes were prepared in two different reducing conditions, adding to the reconstitution mixture 1 or 50 mM DTE. Thus, the antiport rate was measured incubating the reconstituted protein with test molecules at 10 μM final concentration together with 0.1 mM [3H]-carnitine and then stopping the transport activity after 30 min by NEM. The values are means±SD from three independent experiments, significantly different from the controls, as calculated from Student's t-test analysis (* p<0.01).

27279_cbic202000268-fig-5001.jpg

Scheme 1Open in figure viewerPowerPointBiantennary N-glycan with unmodified bisecting GlcNAc (A); complex N-glycans found on human serum IgG bearing a β1,4-galactosylated bisecting GlcNAc motif (B, C); hybrid N-glycans from 19 A glycoprotein with a galactosylated bisecting GlcNAc motif (D, E).

27279_cbic202000268-fig-5002.jpg

Scheme 2Open in figure viewerPowerPointSynthesis of N-glycan 3, deprotection to hemiacetal 4 and enzymatic galactosylation to 5 and 6. The numbering of the residues for NMR assignments is shown for 6.

27279_cbic202000268-fig-5003.jpg

Scheme 3Open in figure viewerPowerPointThe panel of bisected N-glycan azides (7–11) was enzymatically galactosylated under dilute (1.) and subsequently under stringent (2.) conditions. Transfer of an additional galactose to the bisecting GlcNAc was observed only in the case of the biantennary N-glycans 17 and 18. (n. d.=not detected).

27279_cbic202000268-fig-5004.jpg

Scheme 4Open in figure viewerPowerPointSynthesis of the trigalactosylated bisected N-glycan azide 17, conjugation with lanthanide binding tag 18, deprotection and formation of the diamagnetic complex 20 a and the paramagnetic complex 20 b.

27279_cbic202000268-fig-5005.jpg

Scheme 5Open in figure viewerPowerPointA), B) HSQC sections of the anomeric region of the diamagnetic (20 a) and paramagnetic complex (20 b); the three galactose signals are within the dotted lines. C) PCS of paramagnetic complex 20 b. D) Relative percentages of the main conformations calculated from PCS.

12358_celc202101434-fig-5001.jpg

Scheme 1Open in figure viewerPowerPointElectrografting of alkyl silyl protected ethynyl aryl diazonium salts to graphite electrodes.

Graphite electrodes were used as substrates for electrografting experiments according to a method displayed in Scheme #celc202101434-fig-5001#1. The electrografted electrodes are referred to as TMS and TIPS, whereas the deprotected electrodes are labelled as ethynyl. To distinguish the ethynyl groups from each other, the protecting group from which they derived is added in brackets, giving ethynyl (TMS) and ethynyl (TIPS). Silicon in alkyl silyl protecting groups works as a marker to confirm successful electrografting via XPS (Figure #celc202101434-fig-0001#1a). The appearance of Si 2p peaks at a binding energy of 100.5 eV (Si 2p3/2) in TMS and TIPS corresponding to silicon linked to carbon suggests that the electrografting of the protected ethynyl moiety was performed successfully. The silicon content for TMS (3.2 at %) is higher than for TIPS (1.4 at %), which is attributed to the higher distance between the molecules caused by the more steric demanding TIPS group. As expected, Si 2p peaks in ethynyl (TMS) and ethynyl (TIPS) disappeared after the deprotection of the ethynyl moiety. The SEM images of the reference and all modified electrodes (Figure #celc202101434-fig-0001#1b) show conductive additive particles and a fibrous structure on the graphite surface of the pristine, TMS and TIPS samples, which is assigned to the PVdF binder. However, in the images of ethynyl (TMS) and ethynyl (TIPS) additional spots are visible on the electrode surface, whereas ethynyl (TMS) shows visibly more spots than ethynyl (TIPS). The spots most probably evolve during the deprotection process. To investigate the electrochemical performance of ethynyl (TMS) and ethynyl (TIPS) compared to the pristine electrode, a three-electrode setup was used. The electrochemical cycling performance and Coulombic efficiencies are presented in Figures #celc202101434-fig-0001#1c and d, respectively. The delithiation capacities decrease for the modified samples at low rate, whereas constant capacity values are observed for the pristine graphite electrode (Figure #celc202101434-fig-0001#1c). A massive drop in capacity from 335 mAh g−1 to 97 mAh g−1 appears for the modified samples when applying a higher current. This drop (71 %) is much higher than for the pristine (28 %), which gives reason to assume slower intercalation kinetics due to blocked intercalation channels for the modified electrodes. Moreover, the ethynyl (TIPS) values at 1 C fluctuate, indicating inhomogeneous SEI formation, which may be attributed to the creation of “pinholes” on the surface as described in literature. In addition, efficiencies in the first cycle of both samples noticeable dropped by 10 % compared to pristine, indicating more irreversible side reactions during the first lithiation (Figure #celc202101434-fig-0001#1d). Although, the difference in efficiencies at C/10 compared to the pristine is not as drastic as in the first cycle, the consumption of lithium-ions due to side reactions is more distinct for the modified samples (Figure #celc202101434-fig-0001#1d, inset). We assume that the formed SEI is not stable, as side reactions proceed even after 5 cycles. However, after two cycles at 1 C the efficiencies are comparable to the pristine electrode. Broadened and to lower potential shifted reduction peaks in the first cycle are also observed for nitrogen-containing amino and nitro groups (Figure #celc202101434-fig-0006#6c) what also intensifies in the 10th cycle (Figure #celc202101434-fig-0006#6d). However, for carboxy groups, the reduction peaks are slightly shifted to higher potentials and still very sharp. Even though those peaks deteriorate in the 10th cycle (Figure #celc202101434-fig-0006#6d), they are more pronounced than those of amino and nitro electrodes. Due to multilayer formation, a high number of functional groups is available on the electrode surface, also it is very likely that the grafted layer is thicker than for the ethynyl groups. The resulting SEI film alters the lithium-ion transport, which intensifies over cycling (Figure #celc202101434-fig-0006#6d). Therefore, the results from Figure #celc202101434-fig-0006#6a–d affirm our assumptions made earlier for the capacity and efficiency decrease in Figure #celc202101434-fig-0001#1c–d and 3a–b.

12358_celc202101434-fig-0001.jpg

Figure 1Open in figure viewerPowerPoint(a) Si 2p spectra and (b) SEM images of pristine, electrografted and deprotected electrodes; (c) electrochemical cycling stability at C/10 for 5 cycles and 1 C for 45 cycles and (d) corresponding Coulombic efficiencies of the pristine and deprotected electrodes.

We investigated whether the spots appearing on the surfaces of ethynyl (TMS) and ethynyl (TIPS) could also contribute to the poor performance. Therefore, a pristine electrode and bare graphite powder were treated with the deprotection agent (TBAF 0.1 M in THF) and characterized by SEM (Figure #celc202101434-fig-0002#2a). Since the surface of treated graphite powder is free from spots but the treated electrode is not (Figure #celc202101434-fig-0002#2a), the spots originate from a decomposition of the PVdF binder. To further investigate this phenomenon, we conducted XPS measurements of pristine PVdF powder and PVdF treated with deprotection agent. Figure #celc202101434-fig-0002#2b shows the C 1s and F 1s XPS spectra of the two corresponding powders. The C 1s regions were deconvoluted into four peaks: C−C/C−H at 285.0 eV, CH2/C−O at 286.6 eV, COO/CHF at 288.5 eV and CF2 at 290.9 eV. CH2 and CF2 refer to the bonds in PVdF, of which the intensities drastically decrease after treatment with TBAF. The peak in the F 1s region at around 687.9 eV represents CF2 in PVdF, which decreased after deprotection. Moreover, two additional peaks at 683.8 eV for fluoride and 686.1 eV for CHF can be observed for the deprotected sample. C 1s and F 1s spectra of the corresponding electrodes can be found in Figure S1. It is notable that during the preparation of the PVdF (TBAF) sample, we observed that the originally white PVdF powder immediately turns to black when immersed in the deprotection agent (Figure S2). This observation and the XPS results confirm our assumption that side reactions occurred when the PVdF containing electrode was treated with deprotection agent. This is in line with the electrochemical results demonstrated in Figure #celc202101434-fig-0002#2c–d. The graphite electrode, which was solely treated with TBAF exposes a decrease of delithiation capacity and efficiency in the first cycle in comparison to the pristine electrode. These observations point out that a treatment with TBAF causes a visible change of the electrode surface and decomposition of the binder, resulting in poor electrochemical performance. The effect of binder decomposition could likely affect the impact of the grafted surface groups. To avoid side reactions with PVdF, all experiments were repeated using CMC/SBR instead of PVdF. However, SEM and XPS reveal decomposition of CMC binder most likely due to deacylation after treatment with deprotection agent, which negatively influences the electrochemical performance of the electrodes as well (Figures S3–S4). Since the role of the binder is to ensure good particle-particle cohesion and particle-current collector adhesion to enable stable cycling, it is not surprising that binder decomposition results in poor electrochemical performance. Nonetheless, capacities of ethynyl (TMS) and ethynyl (TIPS) are even lower compared to the TBAF treated pristine electrode, which proposes an additional influence of the electrografting process and/or the ethynyl functionality. Therefore, additional functional groups were investigated.

12358_celc202101434-fig-0002.jpg

Figure 2Open in figure viewerPowerPoint(a) SEM images of pristine and with TBAF treated graphite electrode and graphite powder; (b) C 1s and F1s spectra of pristine and with TBAF treated PVdF binder; (c) electrochemical cycling stability at C/10 for 5 cycles and 1 C for 45 cycles and (d) corresponding Coulombic efficiencies of pristine and with TBAF treated graphite reference electrode.

To avoid binder decomposition by a deprotecting agent, ADS with amino, carboxy and nitro moieties were electrografted to graphite electrodes (Scheme #celc202101434-fig-5002#2). The electrografted samples are referred to as amino, carboxy and nitro. In this case, no secondary deprotection step is needed after grafting. Surface analyses via XPS and SEM are displayed in Figure S5. The characteristic N 1s peaks for amino and nitro groups as well as O 1s peaks for carboxy groups confirm successful grafting of the salts. Looking at the influence of these groups on the electrochemical behavior, it can be seen that all electrografted samples show reduced delithiation capacities, especially at a higher current (Figure #celc202101434-fig-0003#3a). Even though the electrodes were not treated with an additional deprotecting agent, the capacities are not improved compared to Ethynyl (TMS) and Ethynyl (TIPS). A reasonable explanation for this behaviour is the radical mechanism of electrografting. Due to the absence of a protecting group, the formation of dense multilayers is more likely, which may block the graphite surface and hinders lithium-ion intercalation. The thick layer could also inhibit electrolyte penetration and change the porosity of the electrode and therefore lead to lower capacities. The initial capacity loss for carboxy and amino are close to the pristine electrode, whereas it increases for nitro (Figure #celc202101434-fig-0003#3b). The efficiencies of amino in the following cycles are similar to the pristine and nitro. Especially carboxy shows reduced efficiencies at C/10. After the current change, amino and nitro show efficiencies similar to the pristine, whereas for carboxy the efficiencies are still below the pristine after 10 cycles. The addition of functional groups via electrografting is expected to lead to a more reactive surface, which would promote more side reactions upon cycling, especially for reducible functional groups.

12358_celc202101434-fig-5002.jpg

Scheme 2Open in figure viewerPowerPointElectrografting of amino, carboxy and nitro aryl diazonium salts to graphite electrodes.

To preserve free intercalation channels, another modification method was tested. Functionalized aryl anilines were mixed with diazotization reagent and graphite powder in an acidic aqueous solution. Since the corresponding ADS are in situ formed and grafted to the graphite powder, this method is referred to as in situ grafting (Scheme #celc202101434-fig-5003#3). Since for in situ grafting the graphite powder is modified before the preparation of the electrode, deprotection of TMS can be done without decomposition of the binder. The resulting electrodes are referred to as TMS, ethynyl (TMS), amino, carboxy and nitro. Again, the appearance of a Si 2p peak for TMS and subsequent disappearance after deprotection confirms successful grafting via the in situ method (Figure #celc202101434-fig-0004#4a). For the amino and nitro groups, nitrogen works as marker molecule. The N 1s peak of amino which is observed after in situ grafting was fitted with two components of −N< at 399.3 eV and N= at 400.7 eV. For the Nitro sample, an additional N 1s peak at 405.7 eV corresponding to −NO2 appears as expected. For Carboxy, both O 1s spectra were deconvoluted by a peak of COO at lower binding energy and a peak of C−O at higher binding energy. Despite the shift in binding energy due to the charging effect, the intensity of COO and C−O peaks increased after in situ grafting, indicating that the electrode was successfully modified. None of the in situ grafted samples shows the fiber-like structure of PVdF in the SEM images (Figure #celc202101434-fig-0004#4b). Carboxy shows a morphology similar to the pristine, whereas ethynyl (TMS), nitro and amino look like they are covered by a film, which supports that multilayer formation also occurs by using this method (Figure #celc202101434-fig-0004#4b). However, the spots observed for electrografted ethynyl (TMS) do not appear for in situ grafted ethynyl (TMS), which confirms the side reactions of TBAF with PVdF.

12358_celc202101434-fig-0003.jpg

Figure 3Open in figure viewerPowerPoint(a) Electrochemical cycling stability at C/10 for 5 cycles and 1 C for 45 cycles and (b) corresponding Coulombic efficiencies of pristine and electrografted amino, carboxy and nitro electrodes.

Compared to the electrografted samples, in situ grafted samples, except amino, show higher delithiation capacities, (Figure #celc202101434-fig-0005#5a). Amino shows a drastic capacity decrease and delivers almost no capacity at higher current, hence, the Coulombic efficiency of >100 % is not meaningful, as the material is not electrochemically active anymore and the charge transfer cannot be attributed to electrochemical storage processes. The very low capacity (less than 13 mAh g−1) stems from capacitive storage which in this case is not regular. Amino groups being activating substituents may cause different grafting behaviour and different electrochemical performance. The capacity of the carboxy sample is comparable to the pristine sample. Ethynyl (TMS) has still a higher capacity (245 mAh g−1) and cycling stability than the pristine (238 mAh g−1) after ten cycles. This trend is preserved up to 45 cycles with a capacity retention of 84 % for ethynyl (TMS) compared to 70 % for the pristine. This means that fewer lithium-ions are consumed for SEI formation and are therefore further available for (de)intercalation. However, at C/10 the capacities drop after the first cycle, due to an increase of side reactions. Even though the efficiencies in the first cycle are higher than for the pristine, this trend changes in the following cycles at low current (Figure #celc202101434-fig-0005#5b). This was also noted for electrografted samples and attributed to increased reactivity of the surface due to an incorporation of functional groups.

12358_celc202101434-fig-5003.jpg

Scheme 3Open in figure viewerPowerPointIn situ grafting of amino, carboxy, nitro and ethynyl (TMS) aryl diazonium salts to graphite electrodes.

SEI formation includes the reduction of electrolyte components and subsequent precipitation of decomposition products on the electrode's surface. In situ grafted samples show the presence of the ethylene carbonate (EC) reduction peak at ∼0.8 V (except amino), whereas this peak is suppressed or even absent for electrografted samples (Figure #celc202101434-fig-0006#6a,c,d, insets). Despite the consumption of lithium-ions during EC reduction, the in situ grafting increased the initial Coulombic efficiency for nitro (88 %), carboxy (89 %) and ethynyl (TMS) (90 %) compared to electrografted analogues (82 %, 86 % and 76 %, respectively) and also slightly compared to the pristine graphite (87 %). It is well known that additives like vinylene carbonate (VC) stabilize the SEI due to polymerization effects. It is likely that triple bonds polymerize as well and therefore influence the properties of the SEI, which can be observed for ethynyl (TMS). We expect that the polymerization is not finalized during the low rate cycling, but already sufficiently developed to enhance the performance at high rates. We do not observe an enhanced rate capability for the electrografting process as the groups are probably arranged too dense and therefore inhibit the insertion of lithium-ions. Differential capacity plots of the first cycle (Figure #celc202101434-fig-0006#6a) additionally show that the lithium-ion transport is much more affected by ethynyl (TMS) than ethynyl (TIPS), given that the reduction peaks are broadened and shifted to lower potentials. After removing the protecting group, there is no difference in the chemical structure of ethynyl (TMS) and ethynyl (TIPS), but the amount of ethynyl groups on the surface is higher for ethynyl (TMS). The rigidity of the dense grafted ethynyl groups and the resulting network during reduction seems to alter the lithium-ion transport. This effect is attenuated if the ethynyl groups are grafted less dense as in ethynyl (TIPS). However, the transport during oxidation is also affected by ethynyl (TIPS) in a similar manner as by ethynyl (TMS), which means the ethynyl group and the resulting decomposition products affect lithium-ion transport in general for samples prepared by the described synthesis route. Differential capacity plots of the 10th cycle confirm this (Figure #celc202101434-fig-0006#6b). Broadened and to lower potential shifted reduction peaks in the first cycle are also observed for nitrogen-containing amino and nitro groups (Figure #celc202101434-fig-0006#6c) what also intensifies in the 10th cycle (Figure #celc202101434-fig-0006#6d). However, for carboxy groups, the reduction peaks are slightly shifted to higher potentials and still very sharp. Even though those peaks deteriorate in the 10th cycle (Figure #celc202101434-fig-0006#6d), they are more pronounced than those of amino and nitro electrodes. Due to multilayer formation, a high number of functional groups is available on the electrode surface, also it is very likely that the grafted layer is thicker than for the ethynyl groups. The resulting SEI film alters the lithium-ion transport, which intensifies over cycling (Figure #celc202101434-fig-0006#6d). Therefore, the results from Figure #celc202101434-fig-0006#6a–d affirm our assumptions made earlier for the capacity and efficiency decrease in Figure #celc202101434-fig-0001#1c–d and 3a–b. For in situ grafted samples, all reduction peaks in the first cycle are sharp and shifted to higher potentials, except amino (Figure #celc202101434-fig-0006#6e). The reduction peaks of amino are broadened and shifted to lower potentials in the first cycle, whereas in the 10th cycle just a flat line is observed, since amino does not deliver any capacity at this point anymore (Figure #celc202101434-fig-0006#6f). For the other samples even at the 10th cycle, the reduction peaks are preserved unlike for electrografted samples. Lithium-ion transport is even enhanced for ethynyl (TMS), given the sharp peaks which are shifted to lower/higher potentials during reduction/oxidation, respectively. Although, in situ grafting does not prevent multilayer formation, the grafted layer is not as dense and thick as it is for electrografted samples, which seems to have a positive impact on lithium-ion transport. However, the performance of carboxy and nitro electrodes still is inferior to the pristine at low and high current. Additional surface analysis is needed for a deeper understanding of the effect originating from the in situ grafted functional groups. However, Figure #celc202101434-fig-0006#6 reveals that the more the EC reduction peak is suppressed, the more the lithium-ion transport is negatively affected at low and high rate. This concerns especially the electrografted samples, whereas the in situ grafted samples show defined EC reduction peaks and better lithium-ion transport at low and high rate. These findings correlate with the observed capacity values of all samples.

12358_celc202101434-fig-0004.jpg

Figure 4Open in figure viewerPowerPoint(a) N 1s, O 1s and Si 2p spectra and (b) SEM images of pristine, in situ grafted (and deprotected) electrodes.

12358_celc202101434-fig-0005.jpg

Figure 5Open in figure viewerPowerPoint(a) Electrochemical cycling stability at C/10 for 5 cycles and 1 C for 45 cycles and (b) corresponding Coulombic efficiencies of pristine and in situ grafted (deprotected) electrodes.

12358_celc202101434-fig-0006.jpg

Figure 6Open in figure viewerPowerPointdQ/dV plots of the first and 10th cycle of pristine and (a), (b) electrografted ethynyl (TMS) and ethynyl (TIPS); (c), (d) electrografted electrodes amino, carboxy, nitro and (e), (f) in situ grafted amino, carboxy, nitro and ethynyl (TMS).

30008_ejic201901100-fig-0001.jpg

Figure 1Open in figure viewerPowerPointPhotograph of fabricated MIL-53(Al) extrudates.

The synthesized and activated MIL-53(Al) powder was extruded with the binder methyl cellulose to yield MOF extrudates (Figure #ejic201901100-fig-0001#1) containing different amounts of binder. The crystallinity of the specimen was verified by powder X-ray diffraction. Figure #ejic201901100-fig-0002#2 (left) depicts the diffraction patterns of the extrudates with 2, 5 and 10 wt.-% of binder MC 400 after fabrication. All specimen show the pattern of the narrow pore MIL-53 phase, which confirms that the crystallinity of the MOF was retained during the extrusion procedure.

30008_ejic201901100-fig-0002.jpg

Figure 2Open in figure viewerPowerPointPowder X-ray diffraction patterns of produced MIL-53 extrudates with different amounts of MC 400 (left). Nitrogen adsorption isotherms at 77 K of the MIL-53 extrudates in comparison to the parent powder (right). Desorption branches are excluded for clarity.

The synthesized and activated MIL-53(Al) powder was extruded with the binder methyl cellulose to yield MOF extrudates (Figure #ejic201901100-fig-0001#1) containing different amounts of binder. The crystallinity of the specimen was verified by powder X-ray diffraction. Figure #ejic201901100-fig-0002#2 (left) depicts the diffraction patterns of the extrudates with 2, 5 and 10 wt.-% of binder MC 400 after fabrication. All specimen show the pattern of the narrow pore MIL-53 phase, which confirms that the crystallinity of the MOF was retained during the extrusion procedure. Figure #ejic201901100-fig-0002#2 (right) shows the nitrogen adsorption isotherms at 77 K of the extrudates MC 400 as well as the isotherm for the MIL-53 powder prior to extrusion (N2 isotherms of extrudates prepared with MC 4000 are shown in the Supporting Information, Figure S1). The N2 isotherm on MIL-53 powder is classified as type I(a) isotherm, which is typical for a microporous materials with a narrow pore size distribution. The calculated BET area and total pore volume of the MIL-53 powder is 1525 m2 g–1 and 0.56 cm3 g–1, respectively, which is in accordance with other MIL-53 powder samples reported before. For all isotherms shown in Figure #ejic201901100-fig-0002#2 (right) no further gas uptake is observed once the saturation plateau is reached. Thus, the shaping procedure developed in this work did not generate voids in the mesoporous regime (2–50 nm) which was previously reported for zeolite beta monoliths prepared with alumina binder. The N2 saturation plateau is reduced with increasing binder content. Furthermore, the isotherm shape now resembles the type I(b) with a more gradual transition from the micropore filling regime to the saturation plateau (Figure #ejic201901100-fig-0003#3). This observation originates from the flexible nature of the framework. In the absence of guest molecules, MIL-53 exhibits a thermally induced phase transition from large to narrow pore between 125 and 150 K., Prior to the first dosage of nitrogen at 77 K, MIL-53 is in its narrow pore form. While the pore opening for the loose MIL-53 powder sample occurs quite fast, the phase transition of MIL-53 extrudates under cryogenic conditions is slowed down by the binder or the compaction of the powder resulting in a shift with respect to gate-opening pressure and a lower isotherm slope (Figure #ejic201901100-fig-0003#3). Since this low pressure regime of approximately 10–3 to 10–1 p/p0 is used to calculate the BET area of microporous materials, this results in an underestimation of the BET area for the shaped pellets. Hence, it is more suitable to compare the accessible pore volumes which are calculated from the adsorption data at higher relative pressures.

30008_ejic201901100-fig-0003.jpg

Figure 3Open in figure viewerPowerPointSemi-log plot of nitrogen adsorption isotherms at 77 K for MIL-53 powder and MIL-53 extrudates containing binder MC 400.

The N2 isotherm on MIL-53 powder is classified as type I(a) isotherm, which is typical for a microporous materials with a narrow pore size distribution. The calculated BET area and total pore volume of the MIL-53 powder is 1525 m2 g–1 and 0.56 cm3 g–1, respectively, which is in accordance with other MIL-53 powder samples reported before. For all isotherms shown in Figure #ejic201901100-fig-0002#2 (right) no further gas uptake is observed once the saturation plateau is reached. Thus, the shaping procedure developed in this work did not generate voids in the mesoporous regime (2–50 nm) which was previously reported for zeolite beta monoliths prepared with alumina binder. The N2 saturation plateau is reduced with increasing binder content. Furthermore, the isotherm shape now resembles the type I(b) with a more gradual transition from the micropore filling regime to the saturation plateau (Figure #ejic201901100-fig-0003#3). This observation originates from the flexible nature of the framework. In the absence of guest molecules, MIL-53 exhibits a thermally induced phase transition from large to narrow pore between 125 and 150 K., Prior to the first dosage of nitrogen at 77 K, MIL-53 is in its narrow pore form. While the pore opening for the loose MIL-53 powder sample occurs quite fast, the phase transition of MIL-53 extrudates under cryogenic conditions is slowed down by the binder or the compaction of the powder resulting in a shift with respect to gate-opening pressure and a lower isotherm slope (Figure #ejic201901100-fig-0003#3). Since this low pressure regime of approximately 10–3 to 10–1 p/p0 is used to calculate the BET area of microporous materials, this results in an underestimation of the BET area for the shaped pellets. Hence, it is more suitable to compare the accessible pore volumes which are calculated from the adsorption data at higher relative pressures. Nitrogen sorption isotherms at 77 K (see Figure #ejic201901100-fig-0003#3) indicated implicitly that all produced MIL-53 extrudates are able to undergo the lp-np phase transition that is well known as the breathing effect. At cryogenic conditions, the breathing is affected by the binder, i.e. the compaction of the crystals into shaped pellets. To investigate the phase transition kinetics at ambient temperature, time-resolved in-situ XRD patterns of the extrudates were collected under defined humid conditions. Figure #ejic201901100-fig-0007#7 exemplary shows the time-resolved XRD patterns for MIL-53 extrudates (5 wt.-% of MC 400) at 45 °C for different relative humidities (RH). The intensity of the reflection corresponding to the (011) plane of MIL-53 large pore phase (MIL-53lp) is decreasing while simultaneously, two reflections attributed to the (200) and (110) planes of MIL-53 narrow pore phase (MIL-53np) arise. Although the phase transition kinetics is slow for low humidity (20 % RH) a complete phase transition is achieved within a few minutes. The kinetics is accelerated at higher humidity. At 45 °C and 45 % RH, MIL-53 extrudates are fully converted from the large pore to narrow pore form within 2.5 min. The MIL-53lp reflection at 8.7° (011) was used to investigate the phase transition kinetics from the large to the narrow pore form. At 20 % RH (Figure #ejic201901100-fig-0008#8 a)), the MIL-53 powder is already in the narrow pore form after 3.7 minutes while all shaped materials require longer times for the transition (roughly 14 minutes). Among the different extrudate batches, no clear relation between the binder content and the kinetics was observed. MIL-53 extrudates with 5 wt.-% of MC 400 reveal a slightly faster pore closing kinetics than the other extrudates with different binder content.

30008_ejic201901100-fig-0004.jpg

Figure 4Open in figure viewerPowerPointSEM images at different magnifications of an MIL-53 extrudate prepared with 5 wt.-% of MC 4000. Methyl cellulose adhesively connects single crystals, but may also cover the outer surface leading to a blocking of pore entrances.

SEM images of the extrudates are shown in Figure #ejic201901100-fig-0004#4 at different magnifications. The upper left image shows the lateral surface as well as the edge of a cylindrical pellet. The outer surface exhibits no macroscopic defects such as cracks or entrapped gas bubbles which would decrease the packing density. The bottom images clearly show that single crystals are connected by nanosized filaments of methyl cellulose via adhesive forces. The voids between the particles are still big enough allowing the MIL-53 crystals to undergo expansion upon breathing as well as providing access to the interior of the extrudate, so that gases can diffuse fast enough to the inner voids of the extrudates. The upper right image of Figure #ejic201901100-fig-0004#4 depicts the case that the binder also partially coats some MOF crystals instead of connecting them. This could possibly block some pore entrances. Figure #ejic201901100-fig-0006#6 illustrates the results of all VCS tests. MIL-53 extrudates containing 2 wt.-% of MC 400 reveal the lowest stability as they already fail at low vertical crushing stress (α = 0.50 MPa). In contrast, the stability is almost doubled when MC 4000 is used. For the MIL-53 extrudates with 5 and 10 wt.-% of MC 400 (Figure #ejic201901100-fig-0006#6 (top)), the Weibull distribution curves coincide. A further increase in binder amount does not improve the mechanical stability. A possible explanation might be that it merely needs a certain, small amount of methyl cellulose to adhesively connect single MOF crystals in the shaped pellet. When too much binder is used, MC may form agglomerates or partially coat the outer surface of the crystals instead of connecting them, as shown in the SEM images (Figure #ejic201901100-fig-0004#4). In consequence, the accessible pore volume is reduced by blocked pore entrances. Surprisingly, the mechanical stability of MIL-53 extrudates with 10 wt.-% of MC 4000 (Figure #ejic201901100-fig-0006#6 (bottom)) is even lower than of extrudates containing 5 wt.-% of MC 4000. Nevertheless, it is assumed that there might be a certain achievable limit for increasing the mechanical stability when methyl cellulose is applied as binder.

30008_ejic201901100-fig-0005.jpg

Figure 5Open in figure viewerPowerPointThermogravimetric analysis in air of MIL-53 powder and three shaped samples MC 400 containing different amounts of binder (2, 5 and 10 wt.-%).

Finally, the produced MIL-53 extrudates were characterized in terms of thermal stability. Thermogravimetric curves of the shaped samples (MC 400 in Figure #ejic201901100-fig-0005#5, TG curves for MC 4000 in the Supporting Information, Figure S2) exhibit three distinct weight loss steps. The first (< 100 °C) and the last one (> 500 °C) correspond to the desorption of the physisorbed water and the collapse of the MIL-53 structure, respectively. The step in-between originates from binder decomposition. Both binders are thermally stable in air up to 250 °C before they decompose at approximately 300 °C. Considering potential applications of MOFs extrudates as adsorbents, e.g. in pressure swing adsorption (PSA) processes, the thermal stability of the binder is high enough to withstand the initial activation of the adsorbent (viz. desorption of water) as well as the temperature rise due to the heat of adsorption. Furthermore, temperature swing adsorption (TSA) may also be taken into account as possible application for MOF extrudates with methyl cellulose as binding agent. The regeneration temperature in TSA processes is in the range of approximately 100–200 °C, which is still at least 50 K below the onset of the decomposition for binder used in this study. It is worth noting that we did not observe any difference in thermal decomposition behavior for the different methyl cellulose binders.

30008_ejic201901100-fig-0006.jpg

Figure 6Open in figure viewerPowerPointEmpirical Weibull distribution curves (open symbols) and fitted Weibull distribution functions (solid lines) of the tested MIL-53 extrudates MC 400 (top) and MC 4000 (bottom) in comparison to the empirical and fitted Weibull distribution curves of a cylindrically shaped carbon molecular sieves (CMS).

Figure #ejic201901100-fig-0006#6 illustrates the results of all VCS tests. MIL-53 extrudates containing 2 wt.-% of MC 400 reveal the lowest stability as they already fail at low vertical crushing stress (α = 0.50 MPa). In contrast, the stability is almost doubled when MC 4000 is used. For the MIL-53 extrudates with 5 and 10 wt.-% of MC 400 (Figure #ejic201901100-fig-0006#6 (top)), the Weibull distribution curves coincide. A further increase in binder amount does not improve the mechanical stability. A possible explanation might be that it merely needs a certain, small amount of methyl cellulose to adhesively connect single MOF crystals in the shaped pellet. When too much binder is used, MC may form agglomerates or partially coat the outer surface of the crystals instead of connecting them, as shown in the SEM images (Figure #ejic201901100-fig-0004#4). In consequence, the accessible pore volume is reduced by blocked pore entrances. Surprisingly, the mechanical stability of MIL-53 extrudates with 10 wt.-% of MC 4000 (Figure #ejic201901100-fig-0006#6 (bottom)) is even lower than of extrudates containing 5 wt.-% of MC 4000. Nevertheless, it is assumed that there might be a certain achievable limit for increasing the mechanical stability when methyl cellulose is applied as binder.

30008_ejic201901100-fig-0007.jpg

Figure 7Open in figure viewerPowerPointTime-resolved XRD patterns of MIL-53 extrudates MC 400, 5 wt.-% binder at 45 °C under humid conditions: 20 %RH (top), 32.5 % RH (middle) and 45 % RH (bottom).

Nitrogen sorption isotherms at 77 K (see Figure #ejic201901100-fig-0003#3) indicated implicitly that all produced MIL-53 extrudates are able to undergo the lp-np phase transition that is well known as the breathing effect. At cryogenic conditions, the breathing is affected by the binder, i.e. the compaction of the crystals into shaped pellets. To investigate the phase transition kinetics at ambient temperature, time-resolved in-situ XRD patterns of the extrudates were collected under defined humid conditions. Figure #ejic201901100-fig-0007#7 exemplary shows the time-resolved XRD patterns for MIL-53 extrudates (5 wt.-% of MC 400) at 45 °C for different relative humidities (RH). The intensity of the reflection corresponding to the (011) plane of MIL-53 large pore phase (MIL-53lp) is decreasing while simultaneously, two reflections attributed to the (200) and (110) planes of MIL-53 narrow pore phase (MIL-53np) arise. Although the phase transition kinetics is slow for low humidity (20 % RH) a complete phase transition is achieved within a few minutes. The kinetics is accelerated at higher humidity. At 45 °C and 45 % RH, MIL-53 extrudates are fully converted from the large pore to narrow pore form within 2.5 min. The MIL-53lp reflection at 8.7° (011) was used to investigate the phase transition kinetics from the large to the narrow pore form. At 20 % RH (Figure #ejic201901100-fig-0008#8 a)), the MIL-53 powder is already in the narrow pore form after 3.7 minutes while all shaped materials require longer times for the transition (roughly 14 minutes). Among the different extrudate batches, no clear relation between the binder content and the kinetics was observed. MIL-53 extrudates with 5 wt.-% of MC 400 reveal a slightly faster pore closing kinetics than the other extrudates with different binder content.

30008_ejic201901100-fig-0008.jpg

Figure 8Open in figure viewerPowerPointPhase transition kinetics at 45 °C and (a) 20 % RH, (b) 32.5 % RH and (c) 45 % RH from large to narrow pore phase for MIL-53 extrudates MC 400 with different binder content in comparison to the parent powder.

Nitrogen sorption isotherms at 77 K (see Figure #ejic201901100-fig-0003#3) indicated implicitly that all produced MIL-53 extrudates are able to undergo the lp-np phase transition that is well known as the breathing effect. At cryogenic conditions, the breathing is affected by the binder, i.e. the compaction of the crystals into shaped pellets. To investigate the phase transition kinetics at ambient temperature, time-resolved in-situ XRD patterns of the extrudates were collected under defined humid conditions. Figure #ejic201901100-fig-0007#7 exemplary shows the time-resolved XRD patterns for MIL-53 extrudates (5 wt.-% of MC 400) at 45 °C for different relative humidities (RH). The intensity of the reflection corresponding to the (011) plane of MIL-53 large pore phase (MIL-53lp) is decreasing while simultaneously, two reflections attributed to the (200) and (110) planes of MIL-53 narrow pore phase (MIL-53np) arise. Although the phase transition kinetics is slow for low humidity (20 % RH) a complete phase transition is achieved within a few minutes. The kinetics is accelerated at higher humidity. At 45 °C and 45 % RH, MIL-53 extrudates are fully converted from the large pore to narrow pore form within 2.5 min. The MIL-53lp reflection at 8.7° (011) was used to investigate the phase transition kinetics from the large to the narrow pore form. At 20 % RH (Figure #ejic201901100-fig-0008#8 a)), the MIL-53 powder is already in the narrow pore form after 3.7 minutes while all shaped materials require longer times for the transition (roughly 14 minutes). Among the different extrudate batches, no clear relation between the binder content and the kinetics was observed. MIL-53 extrudates with 5 wt.-% of MC 400 reveal a slightly faster pore closing kinetics than the other extrudates with different binder content. When the humidity is increased and, thus, the phase transition is accelerated (Figure #ejic201901100-fig-0008#8 b) and c)), all curves tend to converge. There is hardly a noticeable difference left between the kinetics of the powder and the different extrudate batches. It is worth mentioning that the type of binder (MC 400 or MC 4000) had no impact on the breathing kinetics, either (phase transition kinetics for MC 4000 extrudates are given in the Supporting Information, Figure S3). Hence, our shaping procedure does not significantly influence the phase transition kinetics of MIL-53. The phase transition kinetics on shaped flexible metal-organic frameworks are important for the technical applications of flexible MOFs which need to retain the framework dynamics and phase transition kinetics coupled with macroscopic stability, e.g. for methane storage in the flexible Fe(bdp) and Co(bdp) (bdp2– = 1,4-benzenedipyrazolate) frameworks.

30008_ejic201901100-fig-0009.jpg

Figure 9Open in figure viewerPowerPoint(a) Illustration of the large pore and narrow pore structure of MIL-53. (b) CO2 adsorption isotherm of MIL-53 extrudates at 30 °C. The shown data points correspond to the in-situ diffraction patterns (c) at different CO2 pressures. Dashed vertical lines indicate diffraction angle positions of selected reflections characteristic for MIL-53lp (light blue) and MIL-53np (dark blue).

In the previous section, we demonstrated that MIL-53 extrudates undergo the lp-np transition depending on the humidity in their environment. To investigate the framework flexibility in terms of applications related to gas adsorption, we performed in-situ XRD measurements with simultaneous CO2 dosage at 30 °C. In the pressure range from 0 to 25 kPa, MIL-53 stays in its open pore configuration while the CO2 isotherm is almost linear (Figure #ejic201901100-fig-0009#9 (b)). At a CO2 pressure of 28 kPa, reflections indicative of MIL-53 narrow pore phase appear in the diffraction pattern (Figure #ejic201901100-fig-0009#9 c)) while the slope of the CO2 isotherm increases due to enhanced host-guest interactions within the contracting pores. With further increase in pressure, all MIL-53 crystals within the extrudate undergo the phase transition until the MIL-53lp reflections are almost vanished at 50 kPa. At the same time, the slope of the CO2 isotherm decreases again because the narrow pores are filled with adsorbate. Reflections of both, the large pore and the narrow pore form of MIL-53 were identified simultaneously in the pressure range between 25 and 50 kPa (Figure #ejic201901100-fig-0009#9 c)). Thus, there is no uniform threshold pressure inducing the switching from one phase to another but a rather gradual change. Rather than attributing this to the shaping of the powder, we expect that the polydispersity of the crystal size of the synthesized MIL-53 powder is responsible for this observation. Previous reports on a variety of flexible MOFs involving MIL-53(Al)-NH2, [Cu2(bdc)2(npy)n] MOF (bdc = terephthalate, bpy = 4,4-bypyridine), DUT-8(Ni),, and ZIF-8 clearly demonstrated the correlation between crystal size and structural transition pressure. Larger crystals undergo the transformation more easily (for example at lower gate opening pressure) than smaller particles. Since the hydrothermally synthesized MIL-53 powder naturally exhibits a certain distribution in particle size, the “pore closing pressure” regime upon CO2 adsorption broadens.

30008_ejic201901100-fig-0010.jpg

Figure 10Open in figure viewerPowerPointHigh-pressure methane (red) and carbon dioxide (black) isotherms at 30 °C on MIL-53 MC 4000 3 wt.-% of binder (dashed lines) in comparison to the adsorption isotherms on MIL-53 powder (symbols). Closed symbols: adsorption branch, open symbols: desorption branch.

Finally, we measured methane and CO2 isotherms on MIL-53 extrudates in comparison to the MIL-53 powder at 30 °C. Specimens containing 3 wt.-% of MC 4000 were produced and used for the measurements because they exhibit a good tradeoff between mechanical stability and minimal decrease in adsorption capacity (see supporting information). For carbon dioxide, desorption isotherms were also recorded (Figure #ejic201901100-fig-0010#10). Obviously, the sorption capacities of produced extrudates are hardly influenced by the amount of binder added. They are almost similar to the MIL-53 powder (96 % on average in comparison to the powder) and moreover, they are in good agreement with the previously reported isotherm data., Most strikingly, the CO2-induced closing and especially, the pore reopening and hysteresis behavior is not influenced by the binder methyl cellulose. There is no shift in opening pressure, which has been reported for MIL-53 extrudates employing polyvinyl alcohol as binder.

30008_ejic201901100-fig-0011.jpg

Figure 11Open in figure viewerPowerPointHigh-pressure methane (red) and carbon dioxide (black) isotherms at 30 °C on MIL-53-NH2 MC 4000 extrudate with 3 wt.-% of binder. Closed symbols: adsorption branch, open symbols: desorption branch.

In addition, an amino-functionalized MIL-53(Al)-NH2 was synthesized (details on synthesis conditions are given in the Supporting Information) and extruded with 3 wt.-% of MC 4000 into cylindrical pellets. In the absence of guest molecules, MIL-53-NH2 is in the narrow pore form, which is transformed to the large pore form at certain gas pressures. Hence, the breathing behavior of MIL-53-NH2 (np → lp) differs from that of MIL-53 (lp → np → lp). At pressures below 1 MPa, only CO2 is adsorbed within the narrow pores of MIL-53-NH2 extrudates (Figure #ejic201901100-fig-0011#11). Although the kinetic diameters are quite similar (CH4 = 3.76 Å, CO2 = 3.30 Å) solely linear CO2 molecules fit into the narrow pores with dimensions of 2.6 × 13.6 Å2., With increasing pressure, the MIL-53-NH2 extrudates demonstrate their exceptional ability to breathe in the presence of both gases. For CO2, the phase transition starts at 1.3 MPa which is in perfect agreement with the value reported for the MIL-53-NH2 powder sample. In contradiction to the theoretical, and experimental work on the shift in phase transition pressure of embedded soft materials, our examples demonstrate shaping of flexible MOFs with methyl cellulose as binder is possible, leaving the flexible nature of the crystals unaffected.

22995_ejic202100662-fig-0001.jpg

Figure 1Open in figure viewerPowerPointOverview on the cobalt β-diketonate complexes precursors 1–4 used for this work.

Four different Co(II) acetylacetonate complexes (Figure #ejic202100662-fig-0001#1) were used in this work. 1 is commercially available, whereas 2–4 can be prepared by the reaction of CoCl2 with two equivalents Hacac, Htfac or Hhfac in an alkaline aqueous medium at room temperature followed by the slow addition of one equivalent of TMEDA. 2–4 are orange solids and can be purified by sublimation at 120 °C and reduced pressure.

22995_ejic202100662-fig-0002.jpg

Figure 2Open in figure viewerPowerPointIR (a) and UV-VIS measurements (b) of 1–4. The inset in Figure 2a shows enlarged the typical area for water in the IR-spectra for 1. As collected CW EPR spectra at X-band frequency (9.64 GHz) of a) 2, b) 3 and c) 4 at 7 K in black with simulated spectra in red. EPR simulation parameters: 2: geff=[6.52, 3.25, 2.27], A(59Co) =[620, –, –] MHz, linewidth (peak-to-peak)=30 G, HStrain=[500, 1700, 1800] MHz; 3: geff=[6.63, 3.25, 2.18], A(59Co) =[650, –, –] MHz, linewidth (peak-to-peak)=55 G, gStrain=[0.2, 0.8, 0.4]; 4: geff=[4.66, 4.59, 2.32], A(59Co)=[710, 247, –] MHz, linewidth (peak-to-peak)=180 G, gStrain=[0, 0, 0.2] (c). Static TGA curves of 1–4 at 125 °C (d).

The coordination of TMEDA and the substitution of CF3 groups for CH3 groups also lead to noticeable changes in the spectroscopic behaviour of 1–4. Figure #ejic202100662-fig-0002#2a shows IR spectra of these compounds. The band corresponding to the stretching vibration of the C=O group of the acetylacetonate ligand appears around 1600 cm−1. The position of the band shifts to higher wave numbers in the series 1 (1579 cm−1), 2 (1592 cm−1), 3 (1631 cm−1) and 4 (1639 cm−1). This trend is in good agreement with the trend for the C−O-bond length 1 (av. 1.277 Å), 2 (av. 1.263 Å), 3 (av. 1.258 Å) and 4 (av. 1.251 Å). A weak band around 3400 cm−1 for 1 indicates the presence of water, whereas the TMEDA adducts are anhydrous, even after storage at ambient conditions for several weeks. Figure #ejic202100662-fig-0002#2b shows the UV-VIS spectra of the cobalt precursors 1–4. As reference for octahedral coordinated Co2+, an aqueous solution of Co(NO3)2 with the chromophore [Co(H2O)6]2+, which shows the characteristic peak at with a shoulder centred at 510 nm corresponding to typical pink colour of coordinated Co2+ in aqueous solution. The origin of this peak is the 4T1g(F)→4T1g(P) transition. All four compounds show a strong absorption around 290 nm in the ultraviolet range already at low concentrations due to intramolecular π→π* transitions of the acetylacetonate ligand. At higher concentrations, 1–4 also show significant absorption in the visible range, each of which is noticeably splitted. For 1 the signal is centred around 560 nm, while for 2–3 it is clearly blue-shifted and centred around 520 nm. The splitting pattern is similar to the observations on Co(acac)2(EtOH)2 by Pietrzyk et al. who attributed the splitting pattern to the disturbed octahedral geometry. The Continuous-wave (CW) EPR spectra of 2 and 3 at X-band frequency (9.64 GHz) exhibit rhombic g-tensors with resolved hyperfine features along g1. Due to the large separation of the ms=±1/2 and ms=±3/2 states for Co(II) high spin systems, the spectra can be simulated as Seff=1/2 system. The simulations satisfactorily reproduce the features of the spectra to yield a g-tensor with effective g-values of geff=[6.52, 3.25, 2.27] and geff=[6.63, 3.25, 2.18], for 2 and 3 respectively (Figure #ejic202100662-fig-0002#2c). This supports the octahedral nature of the complexes. Multiline patterns for both complexes, associated with A1, can be observed due to the coupling of the unpaired electron with the cobalt nucleus (I(59Co)=7/2, 100 %). The CW EPR spectrum of 4 shows a broad spectrum without resolved hyperfine features. The spectrum can be simulated with an approximate axial g-tensor of geff=[4.66, 4.59, 2.32], indicating an octahedral geometry with axial distortion, caused by the higher fluorination degree.[28.]

22995_ejic202100662-fig-0003.jpg

Figure 3Open in figure viewerPowerPointSEM photographs of CoO_1_Al (a), CoO_2_Al (b), CoO_3_Al (c) and CoO_4_Al (d) grown on Al2O3(0001) with different cobalt precursors at 500 °C.

SEM photographs of the MOCVD grown CoO thin films Al2O3(0001) are displayed in Figure #ejic202100662-fig-0003#3. The film morphology shows a clear dependency with respect to the specific cobalt precursor. While precursor 1 results in the growth of needle-like structures (CoO_1_Al), which probably results from the poor evaporability of 1 and hence the rather low precursor concentration in the gas phase, the use of precursors 2 and 3 leads to the formation of closed films (CoO_2_Al and CoO_3_Al) with rough surface structures consisting of densely packed nanosized grains. Comparable films formed with precursor 4 (CoO_4_Al), but in addition to the nanogranular film, CoO octahedra with well-defined crystal facets and edge lengths in the range of 500 nm are formed.

22995_ejic202100662-fig-0004.jpg

Figure 4Open in figure viewerPowerPointX-ray diffractogram for CoO thin films grown on Al2O3(0001) using different precursors (a), (111)- and (220) pole figures (b and c) for a film grown from 3. The asterix (*) indicates reflections from the Al2O3(0001) substrate.

Figure #ejic202100662-fig-0004#4 shows the Θ-2Θ X-ray diffractograms (XRD) of the cobalt oxide thin films grown on Al2O3(0001) substrates using the cobalt precursors 1–4. Beside the reflections of the substrate, only Bragg reflections corresponding to cubic CoO as the only crystalline phase are visible, while reflexes due to the presence of other crystalline phases, i. e. Co3O4, or metallic Co, crystalline phases were not observed. While the XRD pattern of the CoO film grown from Co(acac)2 shows a polycrystalline nature, the XRD patterns of the films grown from precursors 2–4 exhibit a very strong (111)-preferred orientation, manifested by the very dominant Bragg reflections at 2Θ=36.3° (111) and 77.6° (222). Pole figures were recorded for clarifying the crystallographic relationship between thin film and substrate. In Figure #ejic202100662-fig-0004#4b, the (111)-pole figure of the CoO thin film grown from 3 on Al2O3(0001) recorded at 2Θ=30.5° is displayed. Beside the central peak at ψ=0° and φ=0°, the figure shows six peaks at ψ=70.5° separated by an equal azimuthal distance of 60° in each case. For a perfect 111-orientation a of cubic material such as CoO, only three symmetrically arranged peaks at ψ=70.5° are expected. The appearance of a second set of peaks rotated by 60° against the first set originates from the formation of rotational twins, typically for the growth of materials with cubic crystal structure on substrates with trigonal symmetry. Both sets appear with the same intensity. The twinning can be also observed for the 220-pole figure recorded at 2Θ=61.5° (Figure #ejic202100662-fig-0004#4c). Here six poles are visible at six poles at ψ=70.5° also azimuthal separated by 60°. The poles with three-folded symmetry at ψ=20° originate from the substrate. As visible in Figure #ejic202100662-fig-0004#4c, the diffraction peaks from the CoO film are rotated azimuthal against the peaks from the substrate by 30° which indicates the following epitaxial relationships: CoO(111)//Al2O3(0001) and CoO[−110]//Al2O3[10-10]. X-ray photoelectron spectroscopy (XPS) was performed to gain deeper insight into the elemental composition of the surface and the valence states of the respective elements. XPS confirms the presence of cobalt and oxygen in all deposited films. It is well known that the interpretation of XPS spectra of CoO thin films is complicated due to the metastable nature of the CoO thin film, which leads to the oxidation of the near-surface region and the formation of a thin Co3O4 layer, while the bulk region remains unaffected. Figure #ejic202100662-fig-0005#5a displays the photoemission for the Co 2p region with a complex structure consisting of the Co 2p3/2 und Co 2p1/2 main peaks at 780.3 and 796.2 eV and two satellite peaks at 785.8 and 802.4 eV. This satellite structure is a typical feature of octahedrally coordinated Co2+ in the high-spin state as found in CoO and absent for tetrahedrally coordinated Co2+. The O 1s spectrum (Figure #ejic202100662-fig-0005#5b) consists of two peaks; the peak at lower binding energy (529 eV) is associated with the lattice oxygen from cobalt oxide, whereas the peak higher binding energy at 531 eV can be attributed to the hydroxyl groups of Co(OH)2. Since the intensity of this peak decreases significantly with increasing take-off angle in the angle-dependent XPS measurements, the formation of hydroxides on the surface was caused by the transport and handling of the samples under ambient conditions and is restricted to a very thin near-surface region. The high-resolution C 1s spectra (Figure #ejic202100662-fig-0004#4c), which are comparable for all four samples, can be deconvoluted into two peaks at binding energies associated with C−C bond (285 eV) and C−O bond carbon (288.2 eV). This is the typically signature of adventitious carbon found on all surfaces handled at ambient conditions. The intensity of the C 1s signal decreases strongly, when the XPS up-take angle is decreased. The observed carbon is therefore a near-surface contamination and less due to the thermal decomposition of the acetylacetonate ligand. Since the precursors 2–4 carry nitrogen containing TMEDA ligand, N 1s spectra were also recorded for all samples. These show a weak peak at 400 eV for the different carbon-nitrogen species such as C−NH2. The measured binding energy differs significantly from typical values for metal nitrides, which are usually observed at lower binding energies (∼397 eV). The N 1s shows little dependence on the information depth of the XPS measurement.

22995_ejic202100662-fig-0005.jpg

Figure 5Open in figure viewerPowerPointXPS core level Co 2p (a), O 1s (b), C 1s (c), F 1s (d) and N 1s (e) spectra of films grown on Al2O3(0001) using different precursors.

X-ray photoelectron spectroscopy (XPS) was performed to gain deeper insight into the elemental composition of the surface and the valence states of the respective elements. XPS confirms the presence of cobalt and oxygen in all deposited films. It is well known that the interpretation of XPS spectra of CoO thin films is complicated due to the metastable nature of the CoO thin film, which leads to the oxidation of the near-surface region and the formation of a thin Co3O4 layer, while the bulk region remains unaffected. Figure #ejic202100662-fig-0005#5a displays the photoemission for the Co 2p region with a complex structure consisting of the Co 2p3/2 und Co 2p1/2 main peaks at 780.3 and 796.2 eV and two satellite peaks at 785.8 and 802.4 eV. This satellite structure is a typical feature of octahedrally coordinated Co2+ in the high-spin state as found in CoO and absent for tetrahedrally coordinated Co2+. The O 1s spectrum (Figure #ejic202100662-fig-0005#5b) consists of two peaks; the peak at lower binding energy (529 eV) is associated with the lattice oxygen from cobalt oxide, whereas the peak higher binding energy at 531 eV can be attributed to the hydroxyl groups of Co(OH)2. Since the intensity of this peak decreases significantly with increasing take-off angle in the angle-dependent XPS measurements, the formation of hydroxides on the surface was caused by the transport and handling of the samples under ambient conditions and is restricted to a very thin near-surface region. The high-resolution C 1s spectra (Figure #ejic202100662-fig-0004#4c), which are comparable for all four samples, can be deconvoluted into two peaks at binding energies associated with C−C bond (285 eV) and C−O bond carbon (288.2 eV). This is the typically signature of adventitious carbon found on all surfaces handled at ambient conditions. The intensity of the C 1s signal decreases strongly, when the XPS up-take angle is decreased. The observed carbon is therefore a near-surface contamination and less due to the thermal decomposition of the acetylacetonate ligand. Since the precursors 2–4 carry nitrogen containing TMEDA ligand, N 1s spectra were also recorded for all samples. These show a weak peak at 400 eV for the different carbon-nitrogen species such as C−NH2. The measured binding energy differs significantly from typical values for metal nitrides, which are usually observed at lower binding energies (∼397 eV). The N 1s shows little dependence on the information depth of the XPS measurement.

22995_ejic202100662-fig-0006.jpg

Figure 6Open in figure viewerPowerPointTOF-SIMS depth-profiles for the films CoO_2_Al (a), CoO_3_Al (b) and CoO_4_Al (c).

The angle-resolved XPS measurements confirmed the presence of carbon, nitrogen and fluorine containing species on the sample surface and in the surface-near region (SI, Figure S55). We therefore investigated these sample by time-of-flight secondary ion mass spectrometry (TOF-SIMS). TOF-SIMS depth profiles were recorded using Bi+ ions for all three samples (Figure #ejic202100662-fig-0006#6). Since both nitrogen and fluorine preferentially form anions, the negative mode was chosen for the depth profiles. Due to the high reactivity of the anion and the associated very low secondary ionisation yield, nitrogen is not detected in the form of N− but as a result of a reaction with the carbon contained in the sample as CN−.

22995_ejic202100662-fig-0007.jpg

Figure 7Open in figure viewerPowerPointLSV (a), Overpotentials η10, η25 and η50 (b), EIS (c), Tafel plots (d) and Chronopotentiometry measurements at a current density of 10 mA cm−2 (e) of films grown on platinum.

For the measurement of the electrochemical activity of the CoO films, the MOCVD process was repeated with platinum sheets as substates under the same conditions as described for the grow of CoO on Al2O3(0001), since electrically conductive substrates are required. The MOCVD grown CoO films were evaluated for the OER by testing them as a freestanding working electrode in a standard three-electrode system with 1 M iron-free KOH as electrolyte at room temperature. The working electrodes were activated before the experiments until stable conditions were reached, while linear sweep voltammograms (LSV) with a scan rate of 5 mV ⋅ sec−1 were recorded of the MOCVD grown films and the bare platinum sheets. The overpotential η, which is required for a current density of 10, 25 and 50 mA cm−2, was used to benchmark the electrocatalysts. The LSVs for CoO_2_Pt, CoO_3_Pt, CoO_4_Pt and bare platinum are displayed in Figure #ejic202100662-fig-0007#7a. The CoO thin films on Pt show comparable electrochemical activity for the OER. The overpotential η10 required for a current density of 10 mA ⋅ cm−2 is 440 mV for each of the three films. At higher current densities of 25 and 50 mA cm−2, the overpotential shows comparable values with η25=460 mV and η50=480 mV, respectively. In contrast, the uncoated Pt substrate shows a very low electrochemical activity for the OER, which demonstrates that the electrochemical activity of the CoO films does not originate from the Pt substrate. Figure #ejic202100662-fig-0007#7d shows the Tafel plots for the CoO films, where η is plotted against the logarithmic current density log (j). In addition to the overpotential and Tafel slope, electrochemical impedance spectra (EIS) were recorded for further insight into the electron transfer kinetics. Figure #ejic202100662-fig-0007#7c shows the Nyquist plots at an overpotential of 550 mV. The charge-transfer resistance Rct can be extracted from the diameter of the semi-circular arc. The bare Pt substrate shows a high charge-transfer resistance of Rct=420 Ω. In contrast, the CoO thin film CoO_2_Pt, CoO_3_Pt and CoO_4_Pt show a significant lower Rct of 120, 140 and 160 Ω, respectively.

31323_cmdc201700725-fig-0001.jpg

Figure 1Open in figure viewerPowerPointSchematic representation of the SABRE effect.

An alternative low-cost approach to hyperpolarization uses parahydrogen (p-H2) to create a non-Boltzmann nuclear spin distribution without changing the identity of the molecule of interest. This technique is known as signal amplification by reversible exchange (SABRE) and is shown schematically in Figure #cmdc201700725-fig-0001#1. It can enhance the signals detected by MRI and nuclear magnetic resonance (NMR) spectroscopy across a wide range of nuclei such as 1H, 13C, 15N and others. It works by harnessing the latent polarization of p-H2 through binding to a metal catalyst, typically [Ir(H)2(Sub)3(IMes)]Cl, as hydride ligands. Simultaneous binding of the substrate allows spontaneous transfer of polarization through the scalar coupling network at low magnetic fields. Subsequent substrate dissociation from the catalytic complex allows buildup of hyperpolarized substrate in solution.

31323_cmdc201700725-fig-0002.jpg

Figure 2Open in figure viewerPowerPointEffect of deuterated solvent mixture on cell viability: MTT cell viability assays performed on the indicated cell lines after A) 6 h and B) 24 h of treatment with deuterated solvents at various ratios. The final solvent volume in the cell culture medium was 10 %. EtOD=[D6]ethanol. Data are the mean+SD. *P<0.05, **P<0.005, ns: not significant vs. untreated control (100 % viable); one-way ANOVA.

To assess the toxicity of these solvent mixtures, we performed an appropriate cell viability assay on A549 and MCF7 cells, which were treated with various dilutions of [D6]ethanol in D2O. As shown in Figure #cmdc201700725-fig-0002#2, the viability of both cell lines was significantly decreased if [D6]ethanol (100 %) was added to cell culture medium and treated for a short time (6 h). Conversely, over the same time period (6 h), 50 % [D6]ethanol in D2O (1:1) showed no change in cell viability. Extending the treatment durations to 24 h, however, significantly decreased the viability (Figure #cmdc201700725-fig-0002#2 B). We found that treatment of cells in a 30 % [D6]ethanol in D2O (30:70) solution did not show toxicity to cells over long treatment times relative to other deuterated solvent mixtures (Figure #cmdc201700725-fig-0002#2 B and Supporting Information Figure S1). Together the results indicate that a significant decrease in cell viability is evident when the solvent contained [D6]ethanol concentrations higher than or equal to 50 %. The behavior of analogous protio solvent mixtures is similar (Supporting Information Figure S2). From these in vitro cytotoxicity analyses on SABRE solvents, we conclude that to achieve optimal biocompatibility it is important to consider the duration of exposure on cells in culture (or in vivo) when performing SABRE using solvent with higher (>30 %) [D6]ethanol content.

31323_cmdc201700725-fig-0003.jpg

Figure 3Open in figure viewerPowerPointCytotoxicity of alcohol-solubilized methyl nicotinate: MTT cell viability data showing A), C) A549 and B), D) MCF7 cells treated for 6 h (upper panel) and 24 h (lower panel) with various dilutions of MN and d2-MN solubilized in 30 % [D6]ethanol in D2O. The final solvent volume in the cell culture medium was 10 %. Data are the mean+SD from three independent experiments (n=3). Statistically significant differences from untreated control group (or from protio form of MN) are shown. *P<0.05, ns: not significant vs. untreated control group; one-way ANOVA.

Having found that 30 % [D6]ethanol in D2O does not induce toxicity to cells in vitro, we then performed viability assays for d2-MN dissolved in this solvent mixture. Again, we compared d2-MN with MN to exclude any toxic effects that arise from the selective deuteration in d2-MN when dissolved in alcohol solvents. Importantly, d2-MN shows good solubility in this solvent composition and it did not decrease the viability of A549 and MCF7 cells when treated for up to 6 h (Figure #cmdc201700725-fig-0003#3 A,B, respectively). However, longer treatment times at concentrations higher than 5 mm are shown to affect the viability of both cell lines (Figure #cmdc201700725-fig-0003#3 C,D). Surprisingly, as shown in Figure #cmdc201700725-fig-0003#3 D, when compared with MN, d2-MN induced a significant decrease in the viability of MCF7 cells at this concentration (5 mm) when treated for long time (24 h). This further indicates that deuteration might affect the toxicity effects of the substrate either by itself, or that toxicity is more pronounced as an additive effect when mixed in alcohol solvent at this long exposure time (24 h). Together, our data suggest that the in vitro cytotoxicity of d2-MN in an [D6]ethanol/D2O solvent mixture depends on the duration of exposure on cells in culture. It is worth mentioning that the aim of our cytotoxicity assessment is to allow us to understand the transient effect these compounds or solvents play under the stated conditions. We are aware that in an in vivo setting, biocompatibility would be dependent on physiological status and the pharmacokinetics of the organism and the mechanism of action of the treated material. Nonetheless, the SABRE approach requires the contrast agent to stay in the body for a comparatively short time prior to excretion, as relaxation limits utility and further reduces toxicity concerns.

31323_cmdc201700725-fig-0004.jpg

Figure 4Open in figure viewerPowerPointEvaluating the biocompatibility of SABRE reaction mixture: A) A549 and B) MCF7 cells were treated with various bolus volumes (0, 1.25, 2.5, 5, and 10 %) of SABRE reaction mixture, and cell viability was assessed 1, 6, and 24 h thereafter by MTT assay. Data are the mean+SD from three independent experiments (n=3). *P<0.05, **P<0.005, ns: not significant vs. untreated control (100 % viable); one-way ANOVA.

Given that the cytotoxic dosage of d2-MN in the solvent [D6]ethanol/D2O (30:70) is well above the amounts used for a typical SABRE reaction (considering only less than or equal to 10 % will be taken as a bolus for treatment) we sought to investigate the effect of the SABRE reaction mixture on A549 and MCF7 cell lines. For this, we prepared a typical SABRE solution containing 5 mm of [IrCl(COD)(IMes)] together with 20 mm d2-MN in [D6]ethanol/D2O (30:70) and activated it with 3 bar H2. We exposed the cells to various bolus volumes (1.25, 2.5, 5, and 10 %) of the activated mixture and assessed the viability of the cells at different time periods by MTT assay. As illustrated in Figure #cmdc201700725-fig-0004#4, treatment with the SABRE reaction mixture over a short period of time (1 h) did not decrease the viability of A549 and MCF7 cells when the lowest volume (e.g., 1.25 %) was added to the cell culture medium. However, cells that were treated with 10 % bolus of the SABRE reaction mixture showed less viability at the same time point.

31323_cmdc201700725-fig-0005.jpg

Figure 5Open in figure viewerPowerPointSchematic presentation of the catalyst deactivation procedure: 1. Activation: The active catalyst is formed through reaction of [IrCl(COD)(IMes)], d2-MN, and H2 in [D6]ethanol/D2O (30:70) solution. 2. Deactivation: Addition of BPS leads to immediate formation of the deactivated catalyst [Ir(IMes)(BPS)(d2-MN)(H)2]Cl as confirmed by 1H NMR spectroscopy and LC–MS. 3. Depletion: Ion-exchange chromatography on DEAE-Sephadex® leads to <2 % catalyst contamination.

We have thus developed a protocol to remove the SABRE catalyst from solution. The addition of a chelating ligand to the SABRE reaction prevents reversible exchange of the substrate, deactivating the SABRE process without affecting the polarization levels whilst extending T1 relaxation times. We postulated that the addition of bathophenanthrolinedisulfonic acid disodium salt (BPS) would see it irreversibly bind to the iridium center whilst giving an opportunity to remove the resultant species via ion-exchange chromatography. This procedure is shown schematically in Figure #cmdc201700725-fig-0005#5. First, we prepared the activated SABRE reaction mixture in [D6]ethanol/D2O solution prior to the addition of a solution of 2.0 equiv of BPS in D2O. Following the reaction by 1H NMR spectroscopy reveals the immediate formation of a new hydride species at δ−19.6 ppm which we attribute to [Ir(IMes)(BPS)(d2-MN)(H)2]Cl and confirmed by LC–MS (Supporting Information Figure S7). After filtration through DEAE-Sephadex® with D2O as eluent, less than 2 % of the catalyst remains in solution with high mass recovery of d2-MN, which can be delivered in the biocompatible [D6]ethanol/D2O solvent mixtures. This protocol is therefore efficient at removing the SABRE catalyst from solution. Importantly, the hyperpolarized SABRE signal is still visible after the deactivation and depletion steps. The total signal gains were 74±21-fold which represents a 30 % decrease in signal relative to the standard SABRE sample in [D6]ethanol/D2O (30:70). 1H NMR spectra are shown in the Supporting Information (Figure S7). As the deactivation and depletion process takes a minimum of 12 seconds longer than a standard sample measurement, we attribute the loss to relaxation effects.

31323_cmdc201700725-fig-0006.jpg

Figure 6Open in figure viewerPowerPointAchieving biocompatible SABRE by deactivating the catalyst: MTT viability assay showing A) A549 and B) MCF7 cells treated with various volumes of catalyst-depleted SABRE reaction mixture for 1, 6, and 24 h. Data are the mean+SD from three independent experiments (n=3). *P<0.05, **P<0.005, ns: not significant vs. untreated control (100 % viable); one-way ANOVA.

The cytotoxicity of the catalyst-depleted samples on cells was then evaluated by taking different volumes (1.25, 2.5, 5, and 10 %) of the reconstituted mixture in [D6]ethanol/D20 (30:70) and by following the treatment conditions in the same manner as performed with the non-quenched SABRE reaction mixture. Pleasingly, 10 % of the bolus containing the catalyst-depleted SABRE reaction mixture did not alter the viability of A549 cells for up to 6 h of treatment (Figure #cmdc201700725-fig-0006#6 A). Similarly, MCF7 cells showed no changes in cell viability when treated with higher volumes (10 %) of the catalyst-depleted SABRE mixture for up to 1 h and for up to 6 h with lower volumes (≤5 %, Figure #cmdc201700725-fig-0006#6 B). However, longer treatments (24 h) showed significant decrease at both lower and higher volumes in both cell lines (Figure #cmdc201700725-fig-0006#6 A,B). It is noted that when extrapolating to an in vivo model, it is unlikely to show the same long-term toxicity due to higher metabolic activity and detoxification mechanisms. Together these data indicate that deactivation and removal of the catalyst could overcome the adverse effect of the SABRE reaction mixture when treating live cells under the conditions used here.

31323_cmdc201700725-fig-0007.jpg

Figure 7Open in figure viewerPowerPointEvaluating the biocompatibility of biphasic SABRE reaction mixture: MTT viability assay showing A) A549 and B) MCF7 cells treated with various volumes (0, 2.5, 5, 7.5, and 10 %) of the bolus from the aqueous fraction of a SABRE reaction in biphasic solvents for 1, 6, and 24 h. Data are the mean+SD from three independent experiments (n=3).

We hypothesized that using the recently reported biphasic approach to SABRE catalysis could be a more rapid and facile way to deplete the solution of the iridium catalyst. In this method the SABRE catalyst is located in a chloroform or dichloromethane phase and minimally in the aqueous phase. Conversely, the hyperpolarized substrate is distributed between the two. For toxicity assessment on cells, the aqueous phase was isolated, and various bolus volumes (2.5, 5.0, 7.5, and 10 %) were added to the cell culture medium. We performed the appropriate viability assay at different time points as illustrated in the previous sections. Treatment with the phase-separated SABRE mixture did not decrease the viability of either A549 or MCF7 cells at any of the time points tested (Figure #cmdc201700725-fig-0007#7). The minimal cytotoxic effect observed here is similar to the effect seen when treated with the substrate alone (Supporting Information Figure S9). This result indicates that the cytotoxicity associated with the SABRE reaction mixture is negated by this method. While we are able to produce a biocompatible bolus by this phase-separation method, the polarization level achieved by the biphasic catalysis is approximately 2000-fold for the same substrate (d2-MN), (Supporting Information Figure S10). Nevertheless, we conclude that polarization in the biphasic mixture is higher than that observed in an [D6]ethanol/D2O solution under 3 bar p-H2 and at 298 K.

10292_cctc201901966-fig-5001.jpg

Scheme 1Open in figure viewerPowerPointCatalytic conversion of γ-valerolactone to ethyl valerate.

The XRD analysis of reduced catalysts show diffraction signals assigned to Ni in metallic form (Figure #cctc201901966-fig-0001#1) with characteristic peaks at 2θ=44.6, 52.1 and 76.4 (ICDD: 00-004-0850). The W promoted Ni/HZ5 (crystallinity=94.6 %) exhibited broad diffraction lines for Ni0 phases indicating the interaction of W with Ni. This effect seems to be more pronounced in the Mo−Ni/HZ5 sample (crystallinity=96.3 %), where a line broadening of the Ni(111) plane is observed, and three diffraction signals assigned to the metallic, ionic and promoter interacted Ni phases can be identified (Figure #cctc201901966-fig-0001#1, inset). In the absence of promoter, mainly metallic Ni phases were found in the XRD pattern of Ni/HZ5 catalyst (Figure S1). HR-TEM images of the un-promoted, as well as the Cr-, W- and Mo-promoted Ni/HZ5 catalysts are shown in Figure #cctc201901966-fig-0004#4. It was found that the Ni nanoparticles are well dispersed on the surface of HZ5. The presence of Ni species can be concluded from the observed lattice plane distance of 0.201 nm being characteristic of the Ni(111) plane. Interestingly, Cr-modified Ni/HZ5 catalyst did not exhibit lattice planes related to the CrOx phases, which can be due to its amorphous nature as also concluded from XRD analysis (Figure #cctc201901966-fig-0001#1). Furthermore, there are different lattice plane distances other than that of Ni(111) presented in Mo and W promoted Ni/HZ5 catalysts. The lattice plane distance of 0.271 nm can be attributed to the (001) plane of a NiMoO4 phase, whereas for WO2 (010) lattice plane distance of 0.379 nm was determined. These results are therefore in well accordance with the results obtained from XRD and XPS analysis of the catalysts confirming the presence of a mixed-metal oxide phase only in Mo promoted Ni/HZ5 catalysts.

10292_cctc201901966-fig-0001.jpg

Figure 1Open in figure viewerPowerPointPowder XRD analysis of promoted Ni/HZ5 samples: a) Cr, b) Mo and c) W.

The H2-TPR was used to investigate the reduction behaviour of the catalysts (Figure #cctc201901966-fig-0002#2) and the respective H2-uptakes are given in Table 2. At least two major signals were found in all samples. The main signal in Cr−Ni/HZ5 at 382 °C can be assigned to the reduction of NiO significantly interacting with HZ5. A small hump observed at low temperature ∼150 °C is due to the oxidic Cr species reduction originating from the Crx+ species in a higher oxidation state (3>x<6) in the form of isolated chromate species dispersed on the support. In the case of Mo and W promoted samples, the first signal is assigned to reduction of surfacial NiO interacting with the support and the second one present at high temperature can be assigned to reduction of promoted (Mo or W) NiO and/or bulk NiO species strongly interacting with the support material.

10292_cctc201901966-fig-0002.jpg

Figure 2Open in figure viewerPowerPointH2-TPR profiles of promoted Ni/HZ5 samples: a) Cr, b) Mo and c) W.

The strength and distribution of acid sites on the surface of the catalysts were studied using NH3-TPD analysis (Figure #cctc201901966-fig-0003#3). The obtained results revealed three regions depending on the adsorption strength of NH3 at different temperatures viz. weak (<250 °C), moderate (250–500 °C) and strong (>500 °C) acid sites. Both Mo and W promoted samples exhibit peaks in all three types of acidic regions while the Cr-promoted sample shows only weak and moderate acidic sites. Additionally, it is worth noticing that the proportion of moderate-strong acid sites is relatively higher in the Mo−Ni/HZ5 sample compared to the Cr and W modified catalysts.

10292_cctc201901966-fig-0003.jpg

Figure 3Open in figure viewerPowerPointNH3-TPD profiles of a) Cr, b) Mo and c) W promoted Ni/HZ5 catalysts.

HR-TEM images of the un-promoted, as well as the Cr-, W- and Mo-promoted Ni/HZ5 catalysts are shown in Figure #cctc201901966-fig-0004#4. It was found that the Ni nanoparticles are well dispersed on the surface of HZ5. The presence of Ni species can be concluded from the observed lattice plane distance of 0.201 nm being characteristic of the Ni(111) plane. Interestingly, Cr-modified Ni/HZ5 catalyst did not exhibit lattice planes related to the CrOx phases, which can be due to its amorphous nature as also concluded from XRD analysis (Figure #cctc201901966-fig-0001#1). Furthermore, there are different lattice plane distances other than that of Ni(111) presented in Mo and W promoted Ni/HZ5 catalysts. The lattice plane distance of 0.271 nm can be attributed to the (001) plane of a NiMoO4 phase, whereas for WO2 (010) lattice plane distance of 0.379 nm was determined. These results are therefore in well accordance with the results obtained from XRD and XPS analysis of the catalysts confirming the presence of a mixed-metal oxide phase only in Mo promoted Ni/HZ5 catalysts.

10292_cctc201901966-fig-0004.jpg

Figure 4Open in figure viewerPowerPointHR-TEM images of (a,b) Ni/HZ5, (c,d) Cr, (e,f) Mo, and (g,h) W modified Ni/HZ5 catalysts.

The surface characteristics of Cr, Mo and W promoted Ni/HZ5 samples were examined using XPS (Figure #cctc201901966-fig-0005#5). XPS spectra exhibited two types of Ni species present in Cr and W promoted catalysts. The signals at ∼852.4 eV and ∼855.6 eV (with a satellite signal at ∼861.3 eV) are related to Ni species in metallic and ionic forms, respectively. In contrast, the Mo promoted catalyst exhibits an additional signal at ∼854.2 eV (satellite at ∼858.1 eV) assigned to a Ni species interacting with Mo. These results are in good accordance with the XRD results of the Mo−Ni/HZ5 catalyst, in which three types of Ni species have been observed for the Mo promoted Ni/HZ5 catalyst. It should also be noted that the significant proportion of ionic Ni (NiO form) in all the catalysts is not only caused by incomplete reduction but is also related to surface oxidation of Ni during the sample transfer and exposure to ambient atmosphere before XPS analysis. The surface analysis (XPS) showed no significant well resolved peaks for interacting nickel species in the Cr and W promoted Ni/HZ5 samples. This further supports that alloy formation is more apparent in both the bulk and on the surface of the Mo−Ni/HZ5 catalyst compared to the other catalysts. Furthermore, a well resolved O 1s (Figure S2) spectrum is also observed in the Mo−Ni/HZ5 catalyst compared to the Cr and W promoted catalysts.

10292_cctc201901966-fig-0005.jpg

Figure 5Open in figure viewerPowerPointXPS Ni 2p spectra of promoted Ni/HZ5 samples: a) Cr, b) Mo, c) W.

Pyridine adsorption with FTIR spectroscopy has been used to characterize the nature of acid sites (Brønsted and/or Lewis). The pyridine adsorbate spectra (Figure #cctc201901966-fig-0006#6) show significant proportions of both types, pyridine coordinated to Lewis acid sites (LAS) and protonated pyridine (BAS) on all catalysts indicated by the typical bands at ≈1450 cm−1 and ≈1540 cm−1, respectively. A relatively high amount of acid sites has been detected on Mo−Ni/HZ5 which is in accordance with the NH3-TPD results, while the ratio of BAS/LAS is the lowest in Mo−Ni/HZ5 (Table 2). It is also interesting to notice that the LAS band of the Mo−Ni/HZ5 and the Cr−Ni/HZ5 samples exhibit two components at 1445 and 1451 cm−1 indicating the presence of LAS with different strength. Thus, the percentage of sites of lower strength (1445 cm−1) is higher on Mo−Ni/HZ5. Only one, but broad LAS band is observed in the W promoted sample indicating the possible presence of two types of LAS, too.

10292_cctc201901966-fig-0006.jpg

Figure 6Open in figure viewerPowerPointPyridine adsorbate spectra of pre-reduced promoted Ni/HZ5 catalysts recorded at 150 °C.

To study the interaction with GVL under reaction conditions, GVL was directly dosed into the reaction cell which was flushed with 5 % H2/He. The adsorbate spectra of the pre-reduced catalysts measured after 15 min contact with GVL at 250 °C are displayed in Figure #cctc201901966-fig-0007#7. For comparison, also the spectrum of GVL is shown measured under the same conditions without catalyst in the reaction cell.

10292_cctc201901966-fig-0007.jpg

Figure 7Open in figure viewerPowerPointGVL adsorbate spectra of promoter modified Ni/HZ5 catalysts recorded after 15 minutes of contact with GVL at 250 °C. For comparison the spectrum of pure GVL is also shown.

After the pre-adsorption of GVL, ethanol was injected into the IR cell to initiate the reaction of GVL ring opening to EV. The reaction progress was followed for 40 min, while the cell was permanently flushed with 5 % H2/He. The measured spectra in the carbonyl stretching region are exemplarily shown for the Mo−Ni/HZ5 sample in Figure #cctc201901966-fig-0008#8. The respective overall spectra are shown in Figure S5. Immediately, after dosing ethanol, the intensity of the bands of adsorbed GVL decreases and a new band at 1738 cm−1 emerges which stems from a carbonyl vibration of a product formed by the reaction of adsorbed GVL with ethanol. It is most likely that this band is related to the carbonyl stretching vibration of ethyl valerate, because for ethyl propanoate and ethyl butanoate carbonyl bands were found in the same spectral region around 1738 cm−1 and 1744 cm−1, respectively.

10292_cctc201901966-fig-0008.jpg

Figure 8Open in figure viewerPowerPointIn situ FTIR spectra of Mo−Ni/HZ5 measured after GVL adsorption and subsequent ethanol injection at 250 °C under flushing with 5 % H2/He.

Comparing the spectra of all catalysts measured after GVL pre-adsorption and subsequent dosing of ethanol a similar behaviour was observed (Figure #cctc201901966-fig-0009#9, Figure S6). For all catalysts the intensities of the bands of adsorbed GVL decrease and a band at 1738 cm−1, assigned to the carbonyl vibration of ethyl valerate, appears. However, this band is distinctly more intensive on Mo−Ni/HZ5. This suggests a higher activity of this catalyst compared to the other ones, which is in accordance with the catalytic testing results (cf. Table 1).

10292_cctc201901966-fig-0009.jpg

Figure 9Open in figure viewerPowerPointNormalized in situ FTIR spectra of the promoted Ni/HZ5 catalysts measured after GVL adsorption (bottom) and subsequent ethanol injection at 250 °C (top).

10292_cctc201901966-fig-5002.jpg

Scheme 2Open in figure viewerPowerPointProposed reaction pathways for GVL conversion over metal-promoted Ni/HZ5 catalysts.

4185_cctc202101625-fig-5001.jpg

Scheme 1Open in figure viewerPowerPointMulti-enzymatic cascades coupling TR/Trx1 system [oxidizing dithiothreitol (DTT) to (4S,5S)-1,2-dithiane-4,5-diol (1,2-D, D) to reduce NADP+ to NADPH] with (a) ADH for converting acetophenone (ACP) to (R)-1-phenylethyl alcohol (1-PEA) or with (b) cyclohexanone monooxygenase (CHMO) for converting cyclohexanone (CHO) to ϵ-caprolactone (ECL).

The natural flow of electrons in the thioredoxin system is directed from NADPH over Trx1 and TR to a donor enzyme, as known from ribonucleotide reductases. We reversed this order by the deployment of the reducing agent dithiothreitol (DTT) to TR/Trx1 and NADP+ (Scheme #cctc202101625-fig-5001#1). TR/Trx1 was first cloned into E. coli, isolated by affinity chromatography, and characterized with SDS-PAGE (Figure S1) followed by the kinetic analysis. At neutral pH and 40 °C, the reaction proceeds with a specific activity of 1 U mg−1, calculated on the TR. The dependence of specific activity on enzyme concentrations (i. e., different ratios and amounts of TR and Trx1) is given in the supporting information (Figure S2). The pH spectrum of the TR/Trx1 system showed that the specific activity increases up to 3 U mg−1 at pH 10, while it decreases heavily below a pH 7.5 (Figure #cctc202101625-fig-0001#1a). Concerning the reaction temperature, activity increases up to 70 °C, peaking at 6 U mg−1 and remains stable even up to 80 °C showing the robustness of the system at elevated temperatures (Figure #cctc202101625-fig-0001#1b). Particularly important for a potential technical application is the high stability and the high tolerance against organic solvents. While both, FDH and PDH are inactive around 60 °C, the TR/Trx1 system shows high stability even at 80 °C (Figure #cctc202101625-fig-0001#1b). The TbADH is the only described system that can match the TR/Trx1-system in terms of stability. Also the GDH from Sulfolobus sulfataricus shows rather high thermal stability. The TR/Trx1-system exhibits also high stability when exposed to both water miscible and immiscible organic solvents. Most analyzed solvents did not affect the activity after 1 h of incubation. In comparison, FDHs show decreased activities when exposed to organic solvents. For instance, the FDHs from Rhodococcus jostii and Candida boidinii exhibited 10 % residual activity when exposed to 50 % (v/v) acetonitrile. Also TbADH shows rather high tolerance with 40 % relative conversion in 60 % (v/v) acetonitrile. The comparably high overall stability of the TR/Trx1-system makes it generally suitable for challenging reaction conditions. The pH-dependency was determined at pH values between 5.5 and 10.5 in 50 mM MES, TRIS, HEPES or carbonate buffer, depending on the desired pH value (Figure #cctc202101625-fig-0001#1a). The temperature dependence was measured at the respective reaction temperature including the mentioned 3 min pre-incubation (Figure #cctc202101625-fig-0001#1b).

4185_cctc202101625-fig-0001.jpg

Figure 1Open in figure viewerPowerPointInfluence of pH value (a) and reaction temperature (b) on the DTT-dependent reduction of NADP+ by the TR/Trx1-system. Standard activity assay conditions are 40 °C at a pH value of 8.0 (2 μM TR, 0.4 μM Trx1, 20 mM DTT, 1 mM NADP+, 20 mM MgCl2, 50 mM Tris−HCl).

To analyze the thermostability of the enzyme pair, we measured the activity after storage at 70 °C and 40 °C over the course of two days (Figure #cctc202101625-fig-0002#2a). Trx1 shows only a small decrease of activity of 10 % after 48 h at 70 °C. The activity of TR is reduced to 40 % with a calculated half-life time of 35±4 h at 70 °C. Furthermore, the stability towards organic solvents was analyzed by incubation of the enzyme in 50 % (v/v) mixtures with different water-miscible and immiscible solvents. For all evaluated water-immiscible solvents, no reduction of activity was observed after one hour (Figure #cctc202101625-fig-0002#2b). For the water-miscible solvents, acetone and DMSO led to a reduction of the activity while methanol and acetonitrile had no detectable influence (Figure #cctc202101625-fig-0002#2c). For the temperature stability measurements, an aliquot of each enzyme was incubated at 40 °C and 70 °C, each. The control reaction was performed with both TR and Trx1 incubated at 40 °C. For TR and Trx1 thermostability experiment, the respective enzyme incubated at 70 °C was used together with the other enzyme incubated at 40 °C (Figure #cctc202101625-fig-0002#2a). Solvent stability measurements were performed after 1 h of incubation of the reaction mixture with the respective solvent. For water-miscible solvents, the reaction mixtures were prepared with 50 % (v/v) of the respective solvent. After one hour of incubation, activity assays were performed in the presence of the solvent (Figure #cctc202101625-fig-0002#2c). For water-immiscible solvents, 1 mL of the reaction mixture without NADP+ was incubated with 1 mL of the respective solvent in a rotation mixer at 30 rpm. After separation of the phases by centrifugation, the reaction mixture was removed and used for the activity assay (Figure #cctc202101625-fig-0002#2b).

4185_cctc202101625-fig-0002.jpg

Figure 2Open in figure viewerPowerPoint(a) Stability of Trx1 and TR after incubation at 40 °C and 70 °C for 48 h. Solvent stability for (b) water-immiscible and (c) water-miscible solvents after 1 h incubation (2 μM TR, 0.4 μM Trx1, 20 mM DTT, 1 mM NADP+, 20 mM MgCl2, 50 mM Tris−HCl). The difference in control experiments is due to different enzyme batches used and measurement procedures.

The first investigated reaction system was the ADH catalyzed carbonyl reduction with TR/Trx1 cofactor regeneration (Scheme #cctc202101625-fig-5001#1a). This cascade (TR/Trx1_ADH) includes several parameters, such as the concentrations of ACP/DTT and the amount of respective enzymes. Several outputs were used to evaluate the performance of the cascade, namely the product concentration (1-PEA), reaction selectivity (product over side product), and turnover number (TON) towards NADP+. To simplify the system, the cofactor NADP+ was always kept at relatively high levels of 1 mM rather than being changed as another variate. All reaction components were identified and analyzed with gas chromatography (GC) (Figure S4). The preliminary experiments were carried out with half the amount of DTT to ACP, which gave rise to decent production of 1-PEA in short time (5–8 h) followed by a decrease overnight (Table S3, entries 1–3). By increasing the concentration of DTT, the synthesis of 1-PEA especially the long-time production was highly improved (Table 3, entry 1), indicating the necessity of high levels of cosubstrate to accelerate the reaction to the product(s) side. Also, lower (0.01 U) or higher (0.2 U) levels of TR/Trx1 to ADH did not help, implying the importance of setting a suitable ratio between enzymes. To further prove the reliability and practicability of this cascade, several control experiments were performed accompanying the reaction (50 mM ACP, 100 mM DTT, 0.1 U TR/Trx1, 0.5 U ADH) simultaneously, each excluding one parameter. As is shown in Figure #cctc202101625-fig-0003#3a, the reaction highly outperformed over all controls, offering 35 mM 1-PEA. Expectedly, without ADH, no product was formed while the control without TR/Trx1 led to the least amount of 1-PEA probably due to the glycerol from enzyme stock solution serving as cosubstrate for ADH. The deprivation of cofactor NADP+ did not lower the product as much as others did, which may be attributed to the presence of powerful reductant (DTT) and TR/Trx1 as well as the high affinity for NADP(H) of both enzymes. Aside from ADH, the other representative NADPH-dependent enzyme, cyclohexanone monooxygenase (CHMO), has been selected to further evaluate the practicability of the TR/Trx1-system. ϵ-Caprolactone (ECL) is an important precursor for the polymer synthesis. Being a known tool for the synthesis of ECL from Baeyer-Villiger oxidation of cyclohexanone (CHO), CHMO being paired with ADHs has been applied to some success. However, within this CHMO_ADH system, the production of ECL accompanies with the unwanted side-product formation of cyclohexanol (CHL) from the ADH-catalyzed reduction of CHO. What makes it even more challenging is that CHMO severely suffers from the inhibition by CHL and ECL. This bottleneck may be circumvented by the replacement of ADH with our TR/Trx1 recycling system (Scheme #cctc202101625-fig-5001#1b). Similarly, four reaction parameters, the concentrations of CHO & DTT and the amount of TR/Trx1 & CHMO, were investigated while three target responses (concentration of ECL, selectivity of ECL over CHL, TON for NADP+) were chosen to evaluate the system (Figure S6). Preliminary experiments showed (Table S12), that the synthesis of ECL was achieved without the formation of CHL. The control experiments without either enzymes or cofactor were conducted, where no ECL formation was observed (Figure #cctc202101625-fig-0003#3b). During the pre-experiments, neither less DTT nor more amounts of DTT could improve the synthesis of ECL as the best result was from the reaction condition (50 mM CHO, 50 mM DTT, 0.1 U TR/Trx1, 0.2 U CHMO) (Table 3, entry 5). With the help of DoE screening and optimization steps, a wide range of all parameters was explored (Table S13–S20). We observed the best performance the following reactions conditions (20 mM CHO, 50 mM DTT, 0.1 U TR/Trx1, 0.2 U CHMO) (Table 3, entry 6). ECL formation steadily increases up to a final concentration of 9 mM without detectable CHL side product formation (Figure #cctc202101625-fig-0004#4b). The slight decrease of ECL after 24 h can be attributed to the autohydrolysis of ECL. Despite the relative lower production of ECL in the optimization step due to the degenerated activity of CHMO, the full conversion was achieved within the reaction increasing the cosubstrates-to-substrate ratio (5 mM CHO, 75 mM DTT, 0.075 U TR/Trx1, 0.15 U CHMO) (Table 3, entry 7). In addition, the amount of TR/Trx1 can be increased in further studies to accelerate the reaction equilibrium to the product side (Figure S10). Despite its lower overall yield in comparison to the ADH experiments, the TR/Trx1 system proved to be capable of providing electrons for the CHMO lactone synthesis.

4185_cctc202101625-fig-0003.jpg

Figure 3Open in figure viewerPowerPoint(a) Synthesis of 1-PEA in combined system TR/Trx1_ADH (50 mM ACP, 100 mM DTT, 0.1 U TR/Trx1, 0.5 U ADH) and various controls. (b) Synthesis of ECL in the combined system TR/Trx1_CHMO (50 mM CHO, 100 mM DTT, 0.1 U TR/Trx1, 0.2 U CHMO) and various controls.

To optimize the system, reaction parameters were evaluated with Design of Experiments (DoE) to find the significant ones. In both screening and optimization steps, the impact of reaction parameters including concentration of ACP/DTT, the amount of ADH/TR/Trx1, was analyzed in a defined range on three target responses: the concentration of 1-PEA, reaction selectivity, and TON for NADP+. For the screening step, eight experiments were performed within one block (Table S4–S7), from which the reaction No. 8 (20 mM ACP, 100 mM DTT, 0.05 U TR/Trx1, 0.5 U ADH) led to the best result (15 mM 1-PEA, 69 % yield, 100 % selectivity, 15 TON (NADP+)) (Table 3, entry 2). These screening results further proved our hypothesis got from the pre-experiments: excess DTT was necessary. Optimization was continued by conducting 30 experiments employing the same parameters and target responses, all of which were divided into two blocks including six center points (Table S8–S11). The best performance (25 mM 1-PEA, 68 % yield, 100 % selectivity, 25 TON (NADP+)) was achieved in the reaction No. 26 (Table 3, entry 3). Under these conditions (35 mM ACP, 125 mM DTT, 0.075 U TR/Trx1, 0.3 U ADH), all target responses were gradually improved with increasing reaction time (Figure #cctc202101625-fig-0004#4a). Besides, the enantiomeric excess (ee) was≥99 % determined by GC (Figure S5). It suggests that the concentration of ACP/DTT can be finely adjusted between the ratio 1/2 to 1/5 only in case of the maintenance of high level of DTT. Despite the relatively lower concentrations of both enzymes, a minor excess of TR/Trx1 over ADH still can promote the cascade. Keeping the ratio of ACP/DTT at 1/15 certainly lead to the full conversion of ACP to 1-PEA (Table 3, entry 4). Nevertheless, it turns out that the higher amount of the TR/Trx1 and DTT are helpers to achieve higher production in the cascade (Figure S9). In summary, the TR/Trx1 system is capable of providing electrons for the ADH catalyzed carbonyl reduction, even driving the reaction to completion at high cosubstrates concentrations. Aside from ADH, the other representative NADPH-dependent enzyme, cyclohexanone monooxygenase (CHMO), has been selected to further evaluate the practicability of the TR/Trx1-system. ϵ-Caprolactone (ECL) is an important precursor for the polymer synthesis. Being a known tool for the synthesis of ECL from Baeyer-Villiger oxidation of cyclohexanone (CHO), CHMO being paired with ADHs has been applied to some success. However, within this CHMO_ADH system, the production of ECL accompanies with the unwanted side-product formation of cyclohexanol (CHL) from the ADH-catalyzed reduction of CHO. What makes it even more challenging is that CHMO severely suffers from the inhibition by CHL and ECL. This bottleneck may be circumvented by the replacement of ADH with our TR/Trx1 recycling system (Scheme #cctc202101625-fig-5001#1b). Similarly, four reaction parameters, the concentrations of CHO & DTT and the amount of TR/Trx1 & CHMO, were investigated while three target responses (concentration of ECL, selectivity of ECL over CHL, TON for NADP+) were chosen to evaluate the system (Figure S6). Preliminary experiments showed (Table S12), that the synthesis of ECL was achieved without the formation of CHL. The control experiments without either enzymes or cofactor were conducted, where no ECL formation was observed (Figure #cctc202101625-fig-0003#3b). During the pre-experiments, neither less DTT nor more amounts of DTT could improve the synthesis of ECL as the best result was from the reaction condition (50 mM CHO, 50 mM DTT, 0.1 U TR/Trx1, 0.2 U CHMO) (Table 3, entry 5). With the help of DoE screening and optimization steps, a wide range of all parameters was explored (Table S13–S20). We observed the best performance the following reactions conditions (20 mM CHO, 50 mM DTT, 0.1 U TR/Trx1, 0.2 U CHMO) (Table 3, entry 6). ECL formation steadily increases up to a final concentration of 9 mM without detectable CHL side product formation (Figure #cctc202101625-fig-0004#4b). The slight decrease of ECL after 24 h can be attributed to the autohydrolysis of ECL. Despite the relative lower production of ECL in the optimization step due to the degenerated activity of CHMO, the full conversion was achieved within the reaction increasing the cosubstrates-to-substrate ratio (5 mM CHO, 75 mM DTT, 0.075 U TR/Trx1, 0.15 U CHMO) (Table 3, entry 7). In addition, the amount of TR/Trx1 can be increased in further studies to accelerate the reaction equilibrium to the product side (Figure S10). Despite its lower overall yield in comparison to the ADH experiments, the TR/Trx1 system proved to be capable of providing electrons for the CHMO lactone synthesis.

4185_cctc202101625-fig-0004.jpg

Figure 4Open in figure viewerPowerPoint(a) Representative synthesis of 1-PEA in TR/Trx1_ADH (35 mM ACP, 125 mM DTT, 0.075 U TR/Trx1, 0.3 U ADH). (b) Representative synthesis of ECL in TR/Trx1_CHMO (20 mM CHO, 50 mM DTT, 0.1 U TR/Trx1, 0.2 U CHMO).

20635_chem201903327-fig-5001.jpg

Scheme 1Open in figure viewerPowerPointSynthesis of tetra- and hexadecavalent rhamnosylated ABMs. Reagents and conditions: i) Propargyl α-l-rhamnopyranoside,21 CuSO4⋅5 H2O, tris(3-hydroxypropyltriazolylmethyl)amine (THPTA), NaAsc, DMF/phosphate-buffered saline (PBS), RT, 1 h; ii) azidoacetic22 or pentynoic acid succinimide ester,23 diisopropylethylamine (DIPEA), DMF, RT, 1 h (see the Supporting Information).

We next performed enzyme-linked immunosorbent assay (ELISA)-type studies to evaluate the ability of glycoconjugates 3 and 6 to bind natural Abs. Microtiter plates were first coated with a commercial Rha-functionalized poly[N-(2-hydroxyethyl)acrylamide] polymer (PAA-Rha) and incubated with human serum, a source of anti-Rha antibodies, then fluorescent goat anti-human IgM and IgG. Interestingly, this experiment first allowed to confirm the presence of anti-Rha IgM whereas IgG were not detectable in the serum in the tested conditions (Figure S39 in the Supporting Information). This observation is in good agreement with previous studies showing strong ratio variations of circulating antibodies in human sera. A competitive assay was next performed with glycoconjugates 3 and 6 as inhibitors and Rha as the monovalent control (Figure #chem201903327-fig-0001#1).

20635_chem201903327-fig-0001.jpg

Figure 1Open in figure viewerPowerPointCompetitive binding assay with human serum and monovalent Rha (○), tetravalent (▪, 3) and hexadecavalent (•, 6) conjugates. The assay measures the inhibition of the binding of human anti-Rha IgM to Rha coated on microplate wells (PAA-Rha). The binding was revealed with Alexa-Fluor488-labeled goat anti-human IgM. Inhibition percentage versus the logarithm of the Rha conjugate concentration (in mm) is plotted. Data reported are an average of three independent experiments.

We next assessed the recognition potency of the fluorescein isothiocyanate (FITC)-labelled TBMs 9 and 11 by flow cytometry with two melanoma cells, M21 which expresses αvβ3 integrins and M21-L as the negative cell line control (Figure #chem201903327-fig-0002#2). After having confirm the low expression of αvβ3 on M21-L (Figure S40 in the Supporting Information), we incubated a solution (5 μm) of the monovalent and tetravalent compounds 9 and 11 with both cell lines. As observed previously with similar compounds, we confirmed the higher potency of the clustered cRGD than the monovalent peptide to bind to M21 cells (Figures #chem201903327-fig-0002#2 c and d). In addition, only low unspecific binding was observed with M21-L (Figures #chem201903327-fig-0002#2 a and b). Therefore, scaffold 11 was selected as the TBM to be combined with Rha-based ABMs.

20635_chem201903327-fig-5002.jpg

Scheme 2Open in figure viewerPowerPointSynthesis of mono- and tetravalent fluorescent RGD-based conjugates. Reagents and conditions: i) CuSO4⋅5 H2O, THPTA, NaAsc, DMF/PBS, RT, 1 h (see the Supporting Information).

Past studies demonstrated that ARMs must bind simultaneously both the tumor cell surface and the Abs to form the ternary complex, which is required for immune activation. Therefore, to confirm that the association of ABM and TBMs did not alter their recognition properties, we used an in vitro fluorescent-based assay to analyze the binding of our final ARM molecules to the selected cells. Flow cytometry and confocal microscopy by using purified anti-Rha rabbit IgG antibodies and phycoerythrin (PE)-coupled secondary anti-rabbit IgG antibody have thus been performed (Figure #chem201903327-fig-0003#3). To this end, M21 and M21-L cell lines have been successively incubated with various low concentrations of ARMs and these antibodies to prevent the autoinhibition effect that can occur in three-component binding systems. Although no binding was detected with M21-L control cells whatever the ARM used, we indeed observed a fluorescence increase on the M21 cell line overexpressing αvβ3 integrins, implying the formation of a ternary complex with ARM 15 (100 nm) (Figure #chem201903327-fig-0003#3 d). In addition, as determined by ELISA, no binding occurred when a tetravalent ABM is used, which again indicates the importance of hexadecavalent presentation of Rha for the binding with anti-Rha antibodies.

20635_chem201903327-fig-0002.jpg

Figure 2Open in figure viewerPowerPointEvaluation of the binding of the fluorescent TBMs 9 and 11 (5 μm, red) to M21 (high levels of integrins) and M21-L (low levels of integrins) cells by flow cytometry. a) Binding of 9 to M21-L. b) Binding of 11 to M21-L. c) Binding of 9 to M21. d) Binding of 11 to M21. The autofluorescence of the cells is represented with dashed lines.

This result was also confirmed by visualizing the cells from the same experiment under a confocal fluorescence microscope, which clearly allows to observe the localization of the ARM 15 at the cell surface of M21 but not of the M21-L control cell line (Figure #chem201903327-fig-0004#4). This result clearly demonstrates the accessibility of the ARM 15 to endogenous human antibodies.

20635_chem201903327-fig-5003.jpg

Scheme 3Open in figure viewerPowerPointSynthesis of antibody-recruiting molecules. Reagents and conditions: i) Pentynoic acid succinimide ester, DIPEA, DMF, RT, 1 h; ii) CuSO4⋅5 H2O, THPTA, NaAsc, DMF/PBS, RT, 1 h (see the Supporting Information).

Ideally, ARMs should be delivered without pre-immunization due to the weakened immune system of cancer patients which could strongly limits the ARM efficiency. For this reason, one crucial step is the demonstration that ARM molecules could be used with human serum as unique source of antibodies. Therefore, similar binding experiments have been performed with the human serum used for ELISA assays in which we observed the presence of anti-Rha IgM. Flow cytometry experiments first indicated that the ARM 15 only, that is, with the higher number of Rha units, is able to redirect IgM against M21 cells overexpressing integrins whereas M21-L remained unbound (Figure #chem201903327-fig-0005#5).

20635_chem201903327-fig-0003.jpg

Figure 3Open in figure viewerPowerPointEvaluation by flow cytometry of the binding of ARMs 14 and 15 (100 nm) to M21 and M21-L cell lines by using a purified rabbit anti-Rha IgG antibody (10 μg mL−1) and revealed with a PE-coupled anti-rabbit secondary antibody (1:100). a) Binding of 14 to M21-L. b) Binding of 15 to M21-L. c) Binding of 14 to M21. d) Binding of 15 to M21. Controls without the ARM molecules are represented in black.

Although unspecific IgM binding can be observed with negative M21-L cells, the visualization of the binding by confocal fluorescence microscopy confirmed the formation a complex between the M21 tumor cells, the ARM molecule 15, and anti-Rha IgM present in the human serum (Figure #chem201903327-fig-0006#6). It is also important to mention that the control compound S5 displaying Gal instead of Rha (Figure S42 in the Supporting Information) did not show a positive signal, thus confirming that the formation of the ternary complex is specific to both the ABM and TBM and not to the peptide carrier itself.

20635_chem201903327-fig-0004.jpg

Figure 4Open in figure viewerPowerPointConfocal fluorescence microscopy experiment after incubation with or without the ARM 15 and purified rabbit anti-Rha antibody (10 μg mL−1). The binding was revealed with a PE-coupled anti-rabbit secondary antibody (1:100). a) Control experiment without ARM. b) Experiment performed in the presence of the ARM 15 (100 nm).

20635_chem201903327-fig-0005.jpg

Figure 5Open in figure viewerPowerPointEvaluation of the recruitment of anti-Rha antibodies present in human serum by flow cytometry with M21 and M21-L cell lines revealed by binding with Alexa-Fluor488-coupled anti-human IgM antibody (1:400). a) Binding of 14 (100 nm) to M21-L; b) Binding of 15 (100 nm) to M21-L; c) Binding of 14 (100 nm) to M21; d) Binding of 15 (100 nm) to M21. Controls without the ARMs molecules are represented in black.

20635_chem201903327-fig-0006.jpg

Figure 6Open in figure viewerPowerPointConfocal fluorescence microscopy experiment with M21 cell line after incubation with or without the ARM 15 and normal human serum (50 %). The binding was revealed with Alexa-Fluor488-coupled secondary anti-human IgM antibody (1:400). a) Control experiment without ARM; b) Experiment performed with the ARM 15 (100 nm).

11186_cssc201903428-fig-0001.jpg

Figure 1Open in figure viewerPowerPointMicrobial electrosynthesis of chiral alcohols by using resting E. coli: The enantioselective reduction of acetophenone to (R)-1-phenylethanol takes place in the cytoplasm via the alcohol dehydrogenase from Lactobacillus brevis (LbADH) using NADPH. The cytoplasmatic NADPH pool is linked to the cathode by extracellular electron transfer through methyl viologen (MV) as mediator and putatively the heterologous proteins MtrA, STC, and CymA. The factors examined by the design of experiments approach in this study are highlighted in color, namely concentrations of acetophenone in red, MV in blue, and E. coli cells in orange.

The increasing availability of renewable electric energy from intermittent sources such as wind turbines and solar panels is driving the development of Power-to-X technologies. Electrobiotechnology constitutes a platform for Power-to-Chemicals, which stores inexpensive/excess electrical energy in chemical bonds by combining electrochemistry and biotechnology and offers a plethora of applications. Among these, microbial electrosynthesis (MES) that uses microorganisms as bio(electro)catalysts targets the production of fine and bulk chemicals as demonstrated for, for example, acetic acid, 1,3-propanediol, and α-humulene. Recently, we described a universal chassis for the enantioselective MES of chiral alcohols from cheap ketones by using resting Escherichia coli whole-cell biocatalysts (Figure #cssc201903428-fig-0001#1).

11186_cssc201903428-fig-0002.jpg

Figure 2Open in figure viewerPowerPointResponse surface plots illustrating the relationship between the factors concentrations of methyl viologen and acetophenone at two different E. coli JG622_LbADH cell concentrations (5 g L−1 and 10 g L−1) for the responses yield, reaction rate, and total coulombic efficiency. Red: high values, blue: low values. The underlying data set is provided in Table S1.

Response surface methodology allows analysis of the relationships between operating factors and one or more responses. The surface response plots in Figure #cssc201903428-fig-0002#2 show how the factors concentrations of acetophenone and methyl viologen, as well as cell concentration, affect the responses Y, r, and CEt of the MES. It is worth mentioning that the enantiomeric excess (ee) exceeded 99 % in all cases. Y was determined by the highest measured (R)-1-phenylethanol concentration in each experiment (see the Supporting Information, Section A2). The regression analysis of the experimental Y with a non-linear polynomial regression model (see the Supporting Information, Section A6) provides an excellent model fit (R2=0.98) and precise prediction (Q2=0.95). Within the stipulated conditions, Y appears to be exclusively determined by concentrations of acetophenone and cells, not by changes in concentration of methyl viologen. The highest Y of 90.2 % was achieved at the highest cell concentration of 10 g L−1 but at the lowest acetophenone concentration of 5 mm. The negative effect of increasing acetophenone concentration on Y could suggest that too high acetophenone concentration or increasing concentration of the product (R)-1-phenylethanol are limiting. As acetophenone is hydrophobic, damage, for example, to the cell membrane is likely. Based on these response surface plots, we concluded that the methyl viologen concentration was not limiting Y, underlining the effective constant regeneration by the cathode. The response surface plots from the DoE experiments (Figure #cssc201903428-fig-0002#2) suggested a possible negative effect of high concentrations of acetophenone and/or (R)-1-phenylethanol on Y. To mitigate the potentially negative effects of the substrate acetophenone and simultaneously enhance Y, fed-batch experiments were set up in 1 L electrobioreactors. The feeding rate of acetophenone was set in accordance with the observed reaction rate under optimized conditions to 370 μm h−1, to keep the acetophenone concentration in the electrobioreactor low.

11186_cssc201903428-fig-0003.jpg

Figure 3Open in figure viewerPowerPointPredicted (white) and actual (color filled) values of the optimized batch reaction in A) 250 mL H-cell reactors (n=3) and B) in the 1 L electrobioreactor (n=2). Conditions: 7.58 mm acetophenone, 3.26 mm methyl viologen, 8.97 g L−1 E. coli JG622_LbADH, TEA buffer (pH 7.5), −0.7 V. Yield (Y): red; reaction rate (r): blue; coulombic efficiency (CE): yellow; total coulombic efficiency (CEt): green (see the Supporting Information, Section A3 for CE and CEt calculation details). Values are average±standard deviation (±range in the electrobioreactor case), calculated from independent biological replicates.

Utilizing the collected data from the response surface methodology, we predicted optimal reaction conditions to maximize the three examined performance parameters Y, r, and CEt. As industrial processes are often benchmarked on their productivity, special focus was set on optimizing r. The reaction conditions that are optimal for the highest r were predicted by DoE to be 7.58 mm acetophenone, 3.26 mm methyl viologen, and 8.97 g L−1 E. coli cells, which should lead to a r of 342.5 μm h−1, a Y of 98.4 %, and a CEt of 69.7 %. Experiments were performed in triplicate to verify the predicted optimal conditions (Figure #cssc201903428-fig-0004#4 A). Figure #cssc201903428-fig-0003#3 A illustrates the resulting CEt of 68.4±7.3 %, Y of 94.1±6.9 %, and r of 324.4±66.8 μm h−1 in relation to the prediction. The experimentally achieved results match the prediction almost perfectly (only a deviation of 2–6 %), confirming the excellent prediction precision. It needs to be stressed that the DoE optimization resulted in a 2.4-fold increased Y and a 3.9-fold increased r in comparison to the non-optimized conditions. For H-cell reactors, scale-up is limited, but for transfer of MES to industrial production processes a scalable reaction system is necessary that fits into a biotechnological process environment. To this end, we recently introduced an upgrade kit to turn conventional bioreactors into electrobioreactors, which enable standard process engineering and can be scaled systematically. Utilizing the 1 L electrobioreactors as a platform, combined with the results and information derived from the DoE optimization, MES of (R)-1-phenylethanol was performed. Figure #cssc201903428-fig-0004#4 shows the resulting averaged chronoamperograms as well as the concentration profiles of acetophenone and (R)-1-phenylethanol by using the 250 mL H-cell reactors and 1 L electrobioreactors, respectively. The resulting Y, r, CEt, and CE of the MES in the electrobioreactor are shown in Figure #cssc201903428-fig-0003#3 B. It can be seen that Y (100±2.3 %) and r (372.2±62.2 μm h−1) achieved in the electrobioreactors are superior to those obtained in the H-cell reactors (Y=85.0±2.1 % and r=308.0±18.4 μm h−1), which is possibly due to an enhanced mass transfer. When using the optimized conditions, however, the CEt is inferior in the electrobioreactors compared with the H-cell reactors (CEt=17.0±1.5 % vs. CEt=53.9±2.7 %). The drop of CEt to 17 % in the bioelectroreactors was surprising. When comparing the CE only in the interval of product formation (which is of interest for future bioproduction), the H-cell reactors still show a higher CE than the bioelectroreactors (76.7±2.6 % vs. 53.2±7.5 %), but the drop is not as pronounced (please see Supporting Information Section A3 for the definition of CE). The differences in CE and CEt in H-cell reactors is negligible, which is not the case for the electrobioreactors. This is due to higher amounts of oxygen infiltrating into the scaled-up systems (see the Supporting Information, Section A9). Although the H-cell reactors are strictly gastight and oxygen inflow only happens during sampling, the electrobioreactors are not as gastight owing to the overhead space at the top of the reactors, where a high number of sampling/sensor ports are connected, which are needed for process monitoring and future process control. Also, owing to the high vapor pressure of both acetophenone and (R)-1-phenylethanol, the 1 L reactors were run closed, without sparging of nitrogen gas, not even in the overhead space, to minimize substrate and product losses, which in turn amplified the oxygen intrusion problem (see the Supporting Information, Section A9). Nevertheless, the results illustrate the suitability of the electrobioreactor for successful scaling-up MES using resting E. coli cells.

11186_cssc201903428-fig-0004.jpg

Figure 4Open in figure viewerPowerPointChronoamperograms and acetophenone (AcPh) and (R)-1-phenylethanol (PhEtOH) concentrations in A) 250 mL H-cell reactor and B) 1 L electrobioreactor. Conditions: 7.58 mm acetophenone, 3.26 mm methyl viologen, 8.97 g L−1 E. coli JG622 LbADH, TEA buffer (pH 7.5), −0.7 V. Values are average±standard deviation, calculated from three independent biological replicates (n=3).

Utilizing the collected data from the response surface methodology, we predicted optimal reaction conditions to maximize the three examined performance parameters Y, r, and CEt. As industrial processes are often benchmarked on their productivity, special focus was set on optimizing r. The reaction conditions that are optimal for the highest r were predicted by DoE to be 7.58 mm acetophenone, 3.26 mm methyl viologen, and 8.97 g L−1 E. coli cells, which should lead to a r of 342.5 μm h−1, a Y of 98.4 %, and a CEt of 69.7 %. Experiments were performed in triplicate to verify the predicted optimal conditions (Figure #cssc201903428-fig-0004#4 A). Figure #cssc201903428-fig-0003#3 A illustrates the resulting CEt of 68.4±7.3 %, Y of 94.1±6.9 %, and r of 324.4±66.8 μm h−1 in relation to the prediction. The experimentally achieved results match the prediction almost perfectly (only a deviation of 2–6 %), confirming the excellent prediction precision. It needs to be stressed that the DoE optimization resulted in a 2.4-fold increased Y and a 3.9-fold increased r in comparison to the non-optimized conditions. For H-cell reactors, scale-up is limited, but for transfer of MES to industrial production processes a scalable reaction system is necessary that fits into a biotechnological process environment. To this end, we recently introduced an upgrade kit to turn conventional bioreactors into electrobioreactors, which enable standard process engineering and can be scaled systematically. Utilizing the 1 L electrobioreactors as a platform, combined with the results and information derived from the DoE optimization, MES of (R)-1-phenylethanol was performed. Figure #cssc201903428-fig-0004#4 shows the resulting averaged chronoamperograms as well as the concentration profiles of acetophenone and (R)-1-phenylethanol by using the 250 mL H-cell reactors and 1 L electrobioreactors, respectively. The resulting Y, r, CEt, and CE of the MES in the electrobioreactor are shown in Figure #cssc201903428-fig-0003#3 B. It can be seen that Y (100±2.3 %) and r (372.2±62.2 μm h−1) achieved in the electrobioreactors are superior to those obtained in the H-cell reactors (Y=85.0±2.1 % and r=308.0±18.4 μm h−1), which is possibly due to an enhanced mass transfer. When using the optimized conditions, however, the CEt is inferior in the electrobioreactors compared with the H-cell reactors (CEt=17.0±1.5 % vs. CEt=53.9±2.7 %). The drop of CEt to 17 % in the bioelectroreactors was surprising. When comparing the CE only in the interval of product formation (which is of interest for future bioproduction), the H-cell reactors still show a higher CE than the bioelectroreactors (76.7±2.6 % vs. 53.2±7.5 %), but the drop is not as pronounced (please see Supporting Information Section A3 for the definition of CE). The differences in CE and CEt in H-cell reactors is negligible, which is not the case for the electrobioreactors. This is due to higher amounts of oxygen infiltrating into the scaled-up systems (see the Supporting Information, Section A9). Although the H-cell reactors are strictly gastight and oxygen inflow only happens during sampling, the electrobioreactors are not as gastight owing to the overhead space at the top of the reactors, where a high number of sampling/sensor ports are connected, which are needed for process monitoring and future process control. Also, owing to the high vapor pressure of both acetophenone and (R)-1-phenylethanol, the 1 L reactors were run closed, without sparging of nitrogen gas, not even in the overhead space, to minimize substrate and product losses, which in turn amplified the oxygen intrusion problem (see the Supporting Information, Section A9). Nevertheless, the results illustrate the suitability of the electrobioreactor for successful scaling-up MES using resting E. coli cells.

11186_cssc201903428-fig-0005.jpg

Figure 5Open in figure viewerPowerPointA) Chronoamperogram, B) concentration of acetophenone and (R)-1-phenylethanol, as well as C) the concentration of proteins in the supernatant over time during the fed-batch microbial electrosynthesis. Acetophenone starting concentration: 0.37 mm; feeding rate of acetophenone: 370 μm h−1; 3.26 mm methyl viologen, 8.97 g L−1 E. coli JG622_LbADH, TEA buffer (pH 7.5), −0.7 V. Yellow points: experimental substrate mass balance; dotted line: theoretical mass balance based on substrate feeding rate. Values are average±standard deviation, calculated from three independent biological replicates (n=3).

Figure #cssc201903428-fig-0005#5 shows the concentration progress of acetophenone and (R)-1-phenylethanol (Figure #cssc201903428-fig-0005#5 A) as well as the chronoamperogram during the fed-batch MES (Figure #cssc201903428-fig-0005#5 B). Thereby, the r of 303.7±74.9 μm h−1 is slightly lower than the r observed during the batch experiment, which was to be expected as it is limited by the acetophenone feed (370 μm h−1). The production of (R)-1-phenylethanol stagnates after approximately 40 h, resulting in an absolute titer of 12.8±2.0 mm, which is higher in comparison to the 7.6±0.2 mm achieved in the batch process. CEs of the fed-batch and batch MES in the electrobioreactor were 63.3±12.2 % and 53.9±2.7 %, respectively (please also refer to Supporting Information Section A9). The overall results indicate that feeding of acetophenone and hence keeping its concentration low had a positive impact. As already shown above, a reduced concentration of acetophenone improved the productivity. As organic solvents can have a destabilizing effect on cell membranes, the cell membrane of E. coli might get harmed when acetophenone enters the cell, which could eventually lead to cell membrane disruption and thus, cell lysis. Determination of the protein concentration (Figure #cssc201903428-fig-0005#5 C) in the reaction medium suggests complete cell lysis after approximately 40 h, which is in agreement with no observable colony forming units (CFU) on antibiotic-LB-agar plates after this duration (data not shown). To further investigate the reasons for cell lysis and to determine the possible product inhibition, a batch experiment starting with 7.58 mm (R)-1-phenylethanol in addition to the usual 7.58 mm acetophenone, 3.26 mm methyl viologen, and 8.97 g L−1 E. coli cells was performed. The results of these control experiments showed no significant negative impact of (R)-1-phenylethanol formation at this already high concentration (Figure S10). We further confirmed the electrochemical reversibility of the mediator for the used conditions by using cyclic voltammetry (Figure S13). Nevertheless, the overall mass balance, which considers only the conversion of acetophenone into (R)-1-phenylethanol, closes perfectly (100 %) despite that both acetophenone and (R)-1-phenylethanol are highly volatile (Figure #cssc201903428-fig-0005#5 B). Further, the anoxic reaction system, which is sensitive to oxygen, does not seem to be impacted by the addition of the feeding system. This is, for instance, evident from the constant deep-violet coloring, as well as from the chronoamperograms (Figure #cssc201903428-fig-0005#5 A), where the final background current density remained at a similar level (ca. −0.15 mA cm−2). This further underlines the excellent suitability of the electrobioreactor for the purpose of fed-batch MES.

23878_open201600118-fig-0001.jpg

Figure 1Open in figure viewerPowerPointBioactive natural products that contain a 3-methylfuran core.

The 3-methylfuran framework is found in a variety of NPs with diverse biological activities. Of particular interest are menthofuran (monoterpene), furanoeremophilane (sesquiterpene), cacalol, and tanshinone (quinone; Figure #open201600118-fig-0001#1). Moreover, these compounds are plant secondary metabolites, conferring host resistance against plant invaders. Therefore, we predict that the 3-methylfuran scaffold is an important factor in the activity of these molecules, and we expect to find chemical elicitors by synthesizing NP-like libraries containing the 3-methylfuran moiety. Furthermore, coumarin, chalcone, flavone, flavonol, isoflavone and isoquinolinone are six naturally occurring compounds that have exhibited a variety of biological activities. Inspired by these NPs and their inherent biological activities, we designed a divergent synthetic pathway to synthesize 88 NP-like compounds based on a key 3-methylfuran core obtained from available starting material cyclohexane-1,3-diketone (Scheme #open201600118-fig-5001#1). These compounds have been applied in a phenotypic high-throughput screening of β-glucuronidase (GUS) and several new hits have been discovered. As chemical elicitors, these hits can induce resistance in rice to nymphs of the brown planthopper (BPH) Nilaparvata lugens, one of the most important rice insect pests in Asia.

23878_open201600118-fig-5001.jpg

Scheme 1Open in figure viewerPowerPointDiversity-oriented synthesis of 88 compounds based on six natural product frameworks. Reagents and conditions: a) KOH, ethyl 2-chloroacetoacetate, H2O/MeOH (6:1), RT, 5 days, ≈65 %; b) KOH, H2O/MeOH (1:2.5), RT, 6 h, >90 %; c) Cu, pyridine, DEG, 175 °C, 10 h, ≈85 %; d) NaH, ethyl formate, PhMe, 0 °C–RT, 10 h; e) NaH, ethyl acetate, DME, 0–90 °C, 4.5 h; f) NaH, dimethyl carbonate, DME, 0–90 °C, 4.5 h; g) DDQ, PhMe, 120 °C, 6 h, ≈75 % (d+g); h) DDQ, PhMe, 120 °C, 6 h, ≈70 % (e+h); i) DDQ, PhMe, 120 °C, 6 h, ≈75 % (f+i); j) DMF-DMA, DMF, 75 °C, 4 h, 99 %; k) I2, CHCl3, RT, 15 h, 90 %; l) arylboronic acid, Na2CO3, Pd(OAc)2, PEG 10 000, MeOH, 50 °C, 4 h; m) benzaldehyde, NaH, THF, RT, 2 h; n) 25 % aq NaOH, 30 % aq H2O2, THF/MeOH (3:5), 0 °C–RT, 48 h; o) reactive methylene compound, piperidine, EtOH, 80 °C, 4 h; p) DMSO, NaH, PhMe, 80 °C, 2 h, 95 %; q) benzaldehyde, piperidine, PhMe, 120 °C, 3 h; r) MeOH (saturated with NH3), 65 °C, 24 h, 100 %; s) benzaldehyde, piperidine, PhMe, 120 °C, 12 h. DEG=diethylene glycol; DDQ=2,3-dichloro-5,6-dicyano-1,4-benzoquinone; DMF-DMA=N,N-dimethylformamide dimethyl acetal; DME=glycol dimethyl ether; DMF=N,N-dimethylformamide; PEG=polyethylene glycol; THF=tetrahydrofuran; DMSO=dimethylsulfoxide.

To computationally assess the structural diversity of our 3-methylfuran library, the structural features were analyzed in terms of their chemical properties by using principal component analysis (PCA). PCA computes the position of each compound in a two- or three-dimensional coordinate system based on a set of molecular properties, such as physicochemical properties, to simplify the comparison with different sets of compounds. Using PCA, the chemical properties of our 3-methylfuran library were compared with reference sets of 40 top-selling brand name drugs used by Tan (Table S6) and 20 coumarin and flavonoid natural products (Table S7). Twenty physicochemical properties (Table S3) of these compounds were analyzed using a public, web-based tool. These properties represent each compound as a vector in 20-dimensional space. The 20-dimensional vector can be reduced to two-dimensional vectors by an orthogonal transformation and plotted as a scatter plot (Figure #open201600118-fig-0002#2 A–C and Table S5). The first three principal components captured 80.7 % of the dataset variance (Table S4). As seen in Figure #open201600118-fig-0003#3 a, our library accesses the chemical space occupied by top-selling drugs considerably and overlapped with the space of coumarins and flavonoid natural products. This indicated the potential drug-like properties of our library and possibly implied that these compounds overlapped with biologically relevant areas of the chemical space. Examination of the component loadings (Table S5) indicated that the major contributions of principal component one (PC1) are the number of all atoms, molecular surface areas and solvent-accessible surface areas. The lipophilicity (log P), topological polarity surface area and relative hydrophobic surface area are the key factors associated with principal component two (PC2). The principal component three (PC3) is greatly influenced by relative negatively and positively charge surface areas. It is worth noting that in Figure #open201600118-fig-0002#2 B and C, our library is even more distinct from the range of drug structures compared with the analysis in Figure #open201600118-fig-0002#2 A. These suggest that the relative negatively and positively charged surface areas are two important properties that differentiate our libraries from the drug library in Figure #open201600118-fig-0002#2 B and C. The understanding of such analyses might be advantageous for further structural optimization and provides some insight into the planning of future libraries. PCA was performed using the SPSS 20 software package. A total of 20 physicochemical properties (Table S3) were obtained for established reference sets of 40 top-selling brand-name drugs (Table S6) and 20 coumarins and flavonoid natural products (Table S7). The summary of the contribution of each principal component is shown in Table S4, and the component loadings are shown in Table S5. The first three principal components account for 80.7 % of the variance in the dataset and were used to generate Figure #open201600118-fig-0002#2.