A Simple Complex on the Verge of Breakdown: Isolation of the Elusive Cyanoformate Ion

Cyanide Hitches a Ride Cyanide is a by-product of the biosynthesis of ethylene in plants and it has been somewhat puzzling how the ion is safely removed before it can shut down enzymatic pathways by coordination to catalytic iron centers. A proposed mechanism has implicated the cyanoformate ion—essentially, a weak adduct of cyanide and carbon dioxide—as the initial product, although its lifetime was uncertain. Murphy et al. (p. 75; see the Perspective by Alabugin and Mohamed) crystallized this previously elusive adduct and found that its solution-phase stability varies inversely with the dielectric properties of the medium. The results bolster a picture in which the adduct shuttles the cyanide away from the hydrophobic confines of the enzyme before releasing the cyanide into the more polar aqueous surroundings. Characterization of a cyanide–carbon dioxide adduct bolsters its possible role in protecting a plant enzyme from cyanide inhibition. [Also see Perspective by Alabugin and Mohamed] Why does cyanide not react destructively with the proximal iron center at the active site of 1-aminocyclopropane-1-carboxylic acid (ACC) oxidase, an enzyme central to the biosynthesis of ethylene in plants? It has long been postulated that the cyanoformate anion, [NCCO2]–, forms and then decomposes to carbon dioxide and cyanide during this process. We have now isolated and crystallographically characterized this elusive anion as its tetraphenylphosphonium salt. Theoretical calculations show that cyanoformate has a very weak C–C bond and that it is thermodynamically stable only in low dielectric media. Solution stability studies have substantiated the latter result. We propose that cyanoformate shuttles the potentially toxic cyanide away from the low dielectric active site of ACC oxidase before breaking down in the higher dielectric medium of the cell.

ultimately metabolized, primarily by b-cyanoalanine synthase to produce the amino acid b-cyanoalanine (10), coordination to the iron center of ACC oxidase would seem likely beforehand. A recent publication by Dilley et al. (11) revealed that ACC oxidase is activated at cyanide concentrations between 0.1 and 1 mM, whereas above 1 mM the cyanide becomes inhibitory. These authors propose that cyanide is bound as a ligand to Fe(II) in ACC oxidase, which then assists ACC binding. It should be noted, vis-à-vis our later discussion, that in most studies cyanide levels are typically determined from the bulk plant material (12,13) rather than intra-or extracellularly. How then are the higher levels of cyanide removed from the active site of ACC oxidase for safe metabolization? We believe that the generation of cyanoformate, a simple coordination complex of cyanide and CO 2 , plays a vital role in this regard.
Surprisingly, much of the fundamental coordination chemistry of CO 2 , particularly with regard to small ligands and their corresponding reactivity, has yet to be explored. For example, little is known of the reactions between carbon dioxide and halides or pseudohalides (e.g., OH − , CN − , OCN − , SCN − , and N 3 − ), although this gap can be attributed, at least partially, to the instability of these donor-acceptor complexes. There are only two well-characterized families of halide/ pseudo-halide complexes of CO 2 , the more common being bicarbonates, species that form numerous materials stabilized through extensive hydrogen bonding (14). The only other known example is fluorocarbonate, [FCO 2 ] -. This deceptively simple anion, isoelectronic with [NO 3 ] -, might superficially be predicted to be stable, but it has only recently been prepared, and the fluoride affinity of carbon dioxide was calculated to be -111 kJ mol −1 , which "appears to be the lower limit for a reaction" (15). We began our investigation into the possibility of isolating cyanoformate with a preliminary computational study. Table S6 of the supplementary materials includes a full list of the calculated energies for the association reaction of CO 2 and CN -. At all the levels of theory employed, the formation of cyanoformate was found to be thermodynamically favored in vacuo. However, the binding energy of CO 2 was even smaller than that of the aforementioned fluorocarbonate ion. With the computations indicating that a thermodynamically stable, albeit very fragile, species could exist, we set out to try to isolate a derivative. We had recently shown that higher-order zincates could be stabilized and isolated from ion-rich media using tetraphenylphosphonium halide salts (16). Using a similar approach, we exposed a concentrated solution of [PPh 4 ]CN to an atmosphere of CO 2 , and colorless crystals immediately precipitated. The solvent was decanted, and the crystals were isolated and characterized using crystallographic and spectroscopic techniques. Solid-state nuclear magnetic resonance (NMR) of both singly labeled [PPh 4 ][NC 13 CO 2 ] and doubly labeled [PPh 4 ] [N 13 C 13 CO 2 ] isotopologs confirmed formation of the complex, with d(CO 2 ) = 144.4 parts per million (ppm), consistent with the chemical shift observed for the fluorocarbonate ion,~142 ppm (15). The bonding between the cyano and carbonyl groups was also confirmed by direct observation of one bond coupling [ 1 J 13C-13C = 61(T4) Hz] between the 13 C-labeled carboxyl and 13 C-labeled cyano fragments, based on the resolved 13 CO 2 Fig. 1. Simplified reaction scheme for the generation of ethylene, cyanide, and carbon dioxide from ACC, via cyanoformate, at the active site of ACC oxidase.
www.sciencemag.org SCIENCE VOL 344 4 APRIL 2014 signal in the spectrum of the latter compound (see fig. S19). Figure 2 shows the structure of [PPh 4 ][NCCO 2 ] as determined by single-crystal x-ray diffraction (see figs. S32 and S33, tables S11 to S17, and the accompanying supplementary discussion for further details). Isolation of the fragile anion in the solid state was likely aided by the crystal lattice stabilization of the salt. The C-C bond in the anion [1.480(9) Å] is equal to that reported by Drück et al. (17) in the structure of cyanoformamide, NCC(O)NH 2 . The cyanide C-N bond length [1.056 (12) Å] proved to be significantly shorter than that in cyanoformamide [1.141(8) Å]. Investigation of bond lengths reported in the Cambridge Structural Database (CSD) (18) showed that the observed cyanide bond length is reasonable, if relatively short, suggesting the presence of a strong CN bond with only a weak C-C interaction to the CO 2 portion of the adduct. The strengthening of the CN interaction in the cyanide ligand upon coordination to electron-deficient sites has previously been discussed in the literature (19).
The packing of the ions in the structure of [PPh 4 ][NCCO 2 ] in the tetragonal space group I4 is similar to that observed in many other simple [PPh 4 ]X salts, where X = Br, I, SCN, OCN, N(CN) 2 , and similar (20)(21)(22)(23)(24). Viewed down the z axis, the cations and anions form discrete stacked columns. Because of the large size of the cations, the anions can be thought of as occupying the channels formed between the stacked columns of cations. There are few interion interactions in the structure, and those that are present are generally long and presumably weak. Only the interaction C(2)-H(2)...O(1)' is characterized by a separation of less than 2.5 Å, H(2)…O(1)' = 2.37 Å. The lack of strong interactions between the ions likely results in the relatively free rotation of the anion about the 4 axis in the cavity produced by the parallel stacking of the cations down the z axis.
The theoretically predicted metrical parameters of [NCCO 2 ]show very little variation with respect to the employed method and are in good overall agreement with the results from the x-ray structural study, although some small systematic deviations do exist; numeric data are presented in table S4. Heterolytic splitting of the C-C bond in the anion and subsequent analysis of the bonding energy contributions (electrostatic, orbital interactions, and Pauli repulsion) using the Ziegler-Rauk-Morokuma energy-decomposition procedure (25)(26)(27) show that the linear cyanide and bent carbon dioxide fragments interact predominately (~90%) through a s-type donor-acceptor interaction (28). Furthermore, the results show that the fragile nature of the adduct arises not so much because of the intrinsic characteristics of the C-C bond but because of the large energy release (~200 kJ mol −1 ) associated with linearization of the CO 2 fragment (29). The morphologies of the key frontier orbitals, as well as the results from the analysis of the electron localization function of cyanoformate, are both consistent with the view of cyanide acting as a two-electron donor to carbon dioxide (see figs. S29 and S30 and the accompanying supplementary discussion for full details). Hence, the C-C interaction in [NCCO 2 ]is similar to the bonding in adducts of CO 2 with N-heterocyclic carbenes (30)(31)(32)(33), although, to our knowledge, the latter systems have not been subjected to a thorough theoretical treatment.
Ionic carboxylates exhibit two characteristic C-O stretching bands in their infrared (IR) spectra. The higher-energy asymmetric band (n as CO 2 ) typically appears as a very strong and broad band, whereas the lower-energy band (n s CO 2 ) is usually less intense. The spectra collected on samples of [PPh 4 ][NCCO 2 ] (figs. S3 to S5) exhibit both of these expected signals. Consistent with the theoretical predictions for the IR and Raman bands (tables S8 to S10), the cyano stretch (nCN) is not observed in the infrared spectrum but is clearly visible in the Raman (2196 cm −1 ) ( fig. S7). This frequency, when compared with that of other cyano groups, such as in acetonitrile (2253  We also prepared 13 C-labeled samples to confirm our assignments, and these data are included in the supplementary materials. To visualize the effect of cyanide coordinating to CO 2 , we have superimposed maps of the electrostatic potential on the total electron density of the reactant and product ions (Fig. 3). These plots aid in the assessment of reactivity, in particular prediction of potential sites of attack for nucleophiles and electrophiles. In CN -, the surface around the carbon atom is electron poor compared with the nitrogen center, but both atoms have electron-rich areas directly on the bond axis, consistent with the two possible coordination modes for this anion. The electrostatic potential of [NCCO 2 ]shows that it is now the oxygen centers that are electron rich and therefore the most favored sites for electrophilic interactions. This suggests that hydrogen-bond donors and acidic metals will be attracted more favorably to the oxygen atoms rather than to the cyanide portion of the [NCCO 2 ]anion. The delicate interplay of nucleophilic and electrophilic interactions is very important in supporting the many biological reactions involving CO 2 (34), and the chemistry of the cyanoformate anion is no exception in this regard.
We have also calculated the thermodynamics for the formation of cyanoformate as a function of various solvents at different density functional levels of theory (see Fig. 4, fig. S31, table S7, and the accompanying supplementary discussion for full details). The results suggest that [NCCO 2 ]is thermodynamically stable only in nonpolar solvents (toluene); in high-dielectric media, the anion becomes thermodynamically unstable with respect to fragmentation to CNand CO 2 (Fig. 4). However, the anion is kinetically stabilized, albeit not by much, in all media because breaking the C-C bond has an activation barrier of~40 kJ mol −1 at these levels of theory. Although many enzymatic processes occur in bulk water, the local dielectric constant at the active site of an enzyme is typically low (7,35), in the range of 4 to 20, as compared to 78 for water (36,37). In the cell, at the low-dielectric active site of the ACC oxidase enzyme, the cyanide ions could remain complexed with CO 2 in the form of [NCCO 2 ] -, protecting the iron center from attack. However, as the cyanoformate anion migrates out to the bulk aqueous solution of the cytosol, the thermodynamics are modified, and the complex decomposes to the observed products, namely hydrogen cyanide and carbon dioxide.
The pivotal role assigned to the cyanoformate anion, based on its calculated (in)stability in solutions of different dielectric constant, was bolstered by carrying out further decomposition experiments (see the supplementary discussion for full details 2 ]/IL solution was added to the scrupulously dried solvent being studied, and the disappearance of the n as CO 2 peak of cyanoformate (1689 cm −1 ) was monitored using real-time Fourier transform IR (FTIR) spectroscopy. Half-lives were determined from linear regressions of plots of ln(P/P 0 ), the ratio of peak height (P) to initial height peak (P 0 ), versus time (t). Solution halflives for cyanoformate of 110 min, 55 min, and 17 min were calculated for toluene, tetrahydrofuran, and acetonitrile, respectively. As predicted by theory, the stability of the cyanoformate anion in solution rapidly decreased with increasing dielectric constant of the solution.
Although large-scale experimental studies of the hydrolysis of cyanoformate were not feasible because of the possible release of HCN, the theoretical behavior of the cyanoformate anion in aqueous environments was investigated in more detail by performing geometry optimizations (wB97X-D/aug-cc-pVTZ level) that included explicit H 2 O molecules in addition to the bulk solvent (see the supplementary discussion for full details). In all cases, the calculations minimized to hydrogenbonded structures (O⋅⋅⋅H and/or N⋅⋅⋅H) with no indication of spontaneous hydrolysis. Although a concerted hydrolysis of cyanoformate in water thus seems unlikely, a sequential pathway can be envisioned. As shown computationally, cyanoformate readily dissociates into CNand CO 2 in high-dielectric media, and the cyanide anion is an excellent nucleophile that can easily abstract a proton from water (pK a of HCN = 9.24). The hydroxide ions generated can, in turn, attack the electrophilic carbon of CO 2 to yield the bicarbonate anion. Thus, in an aqueous environment containing dissolved CO 2 , the addition of cyanide will lead to the formation of bicarbonate. This is a possible explanation for the observed activation of ACC oxidase by cyanide (11), because bicarbonate is an important participant in the catalytic cycle of the enzyme (6-8).
Our experimental work has clearly shown that cyanoformate is readily converted to bicarbonate (and HCN) in the presence of even a trace of water. Early attempts to isolate the cyanoformate salt often gave crystals of tetraphenylphosphonium bicarbonate, whose structure has been confirmed by crystallographic analysis. In our 13 C NMR study, the presence of a peak at 162.3 ppm is certainly attributable to bicarbonate, because the chemical shift correlates well with that reported for other such salts (38). That the bicarbonate originates from the added CO 2 is supported by the large size of this peak relative to others in the same spectrum, because 13 CO 2 was used in the original synthesis. We also probed the decomposition of the synthesized cyanoformate salts in a protic solvent by monitoring the n as CO 2 stretch of the cyanoformate anion (1689 cm −1 ) with  www.sciencemag.org SCIENCE VOL 344 4 APRIL 2014 real-time FTIR as 1-propanol was added to a dilute [Bu 4 N][O 2 CCN]/IL solution. This peak rapidly diminished during the addition, while concomitantly a peak corresponding to bicarbonate appeared at 1652 cm −1 (39). Over the course of only 2 min, n as CO 2 for cyanoformate disappeared completely, while the bicarbonate peak reached a steady absorbance. All of the observed experimental results are entirely consistent with the computationally based proposal of a stepwise process for the decomposition and hydrolysis of the cyanoformate anion.
The stability of cyanoformate, a simple anion of CNcoordinated to CO 2 , has been shown to be dependent on the dielectric constant of its local environment. We propose that this property allows it to shuttle toxic CNaway from the lowdielectric active site of the enzyme ACC oxidase before its decomposition in higher dielectric media. In broader terms, the ability to manipulate solution stability of otherwise unstable or transient species, through changes to the dielectric constant, should find other important applications. The small and active Saturnian moon Enceladus is one of the primary targets of the Cassini mission. We determined the quadrupole gravity field of Enceladus and its hemispherical asymmetry using Doppler data from three spacecraft flybys. Our results indicate the presence of a negative mass anomaly in the south-polar region, largely compensated by a positive subsurface anomaly compatible with the presence of a regional subsurface sea at depths of 30 to 40 kilometers and extending up to south latitudes of about 50°. The estimated values for the largest quadrupole harmonic coefficients (10 6 J 2 = 5435.2 T 34.9, 10 6 C 22 = 1549.8 T 15.6, 1s) and their ratio ( J 2 /C 22 = 3.51 T 0.05) indicate that the body deviates mildly from hydrostatic equilibrium. The moment of inertia is around 0.335MR 2 , where M is the mass and R is the radius, suggesting a differentiated body with a low-density core.
A fter Titan, Enceladus has been the most observed satellite by the Cassini spacecraft during its exploration of the Saturnian system. Images of this small moon (252 km radius) revealed a large plume ejected from the south-polar region, with the source being long fractures from which vapor and ice emerge as discrete jets (1). Concurrent observations of sodium and potassium salts in the plume (2), and the measured temperatures within the fractures (3), strongly argue for the presence of liquid water in the subsurface source region. The plume exhibits a time variability well correlated to the predicted tidal stresses of the body (4).
The endogenic (nonsolar) power emitted from the south-polar region, derived from Cassini Composite Infrared Spectrometer data, is 15.8 GW, with a 20% formal uncertainty (5). This is equivalent to an average surface heat flux of~20 mW/m 2 and is an order of magnitude larger than conven-tional estimates of tidal heating if Enceladus' current orbital eccentricity represents a so-called "equilibrium" resonant state with other satellites (6). It indicates time-variability in its internal properties (7), in a resonant state with other nearby moons (8), or in the rate of heat transport. In any or all of these cases, a plausible internal structure is that of a liquid water ocean overlain by a (thermally conductive) crust (5).
The design of the Cassini spacecraft does not allow radio tracking from Earth during remotesensing observations. Therefore, only 3 of the 19 flybys of Enceladus completed so far have been used for gravity measurements. In these close encounters, the spacecraft was continuously tracked from ground antennas while flying within 100 km of the moon's surface, twice above the southern