Introduction

Gold complexes are known to exhibit medicinal properties, [1, 2] and their modern clinical use dates back to 1929. Various gold(I) thiolates including sodium aurothiomalate and auranofin, (2,3,4,6-tetra-o-acetyl-β-1-d-glucopyranosato-S)-triethylphosphine gold(I), have been used to treat conditions such as rheumatoid arthritis. Auranofin is now under clinical trials as an anti-cancer drug, and it has also shown great promise for treatment of parasitic diseases [3]. Numerous other gold(III) species exhibiting anti-tumor activity are also currently under investigation as therapeutics. [4, 5]

Gold complexes are conventionally synthesized through solution-phase chemistry, which typically result in lower yield, with a propensity for the reduction of gold(I) or gold(III) to elemental gold and of ligands to their oxidized products unless the reactions are carried out under inert atmosphere. An alternative to solution-based syntheses is to carry out chemical reactions in the solid phase by grinding the reactants together in mortar and pestle—a technique generally known as mechanochemistry. Mechanochemical synthesis is an emerging methodology through which it has been demonstrated that coordination complexes can be easily produced in very high yields, and in some cases, side reactions can be avoided [6, 7]. The reactions in the solid state could be more appropriate for synthesis of metal complexes, which are often more rigorous to synthesize in solution due to rapid redox reactions stemming from the presence of dissolved oxygen. For example, reactions of M(en)(NO3)2 (M=Pd, Pt; en=ethan-1,2-diamine) with 4,4′-bipyridine (bipy) form tetranuclear square [M(bipy)(en)]4(NO3)8] complexes during mechanochemical synthesis [8]. This reaction illustrates the benefits that solid-state reactions offer over those in solution; here the formation of the platinum complex takes 4 weeks at 100 °C in solution, whereas it can be synthesized in ~ 10 min by grinding the reactants together.

However, a major challenge faced by mechanochemical techniques is accurate characterization of the product in the solid state. Only a few analytical techniques, such as solid-state NMR and X-ray diffraction, are available for this purpose. Gold(I) complexes possess few distinctive features in their UV-visible spectra, and 197Au-NMR signals are typically very broad and uninformative. On the other hand, electrospray ionization-mass spectrometry (ESI-MS) and multi-nuclear 1H, 13C, and 31PNMR methods provide useful data regarding the speciation and reactivity of gold compounds [9,10,11,12,13,14]. Selectivity and structural elucidation capabilities of MS methods, specifically tandem MS analyses, proved advantageous for these assays. Furthermore, these species and the environments in which they reside have the potential to be directly probed when utilizing next-generation, ambient ionization techniques [15,16,17], allowing investigation in the native and biomimetic condensed phases.

Besides application in areas such as forensics [18,19,20,21] and biomedical imaging [22], desorption electrospray ionization–mass spectrometry (DESI-MS) has shown the capability to probe native chemical reactions in solution, [23] as demonstrated for reactions in low-volume droplets [24] and for rapid electrochemical reactions in thin films [25, 26]. While appearing specifically useful for these systems, the application of DESI-MS in solid-phase reaction monitoring is relatively less explored, with the exception of monitoring Baeyer-Villiger solid-phase reactions by Xie and co-workers [27]. Application of DESI-MS to the analysis of gold-containing products in solid-phase, mechanochemical reactions present an interesting opportunity. In this work, auranofin, a gold-based drug, and its precursor triethylphosphine gold(I) chloride were reacted in the solid phase with thiol-containing amino acids, utilizing DESI-MS to probe the complex array of cognizant ligand exchange/scrambling reactions.

Experimental Section

Samples and Materials

The chemicals used in this study were the following: acetic acid (Fisher Scientific, Hampton, NH, USA); auranofin (a gift from GlaxoSmithKline, USA); l-cysteine (Millipore Sigma, St. Louis, MO, USA); cysteine methyl ester hydrochloride (Fisher); glutathione reduced (Millipore Sigma); n-acetylcysteine (Fisher); triethylphosphine gold(I) chloride (Millipore Sigma). Deionized water (resistivity ≈ 18.0 MΩcm) generated by a Barnstead Nanopure® system and HPLC-grade methanol (Millipore Sigma) were used for all DESI-MS spray solvents. Some materials utilized in the study were abbreviated for brevity, as follows: auranofin: Et3PAuStagl; l-cysteine: CySH; o-methylcysteine: o-MeCySH; n-acetylcysteine: n-AcCySH; glutathione: GSH; triethylphosphine-gold(I) chloride: Et3PAuCl.

Mechanochemical Syntheses

Accurately measured equimolar quantities of the gold compound and the target thiol were mixed and ground vigorously with an agate mortar and pestle for 7–10 min. Product mixtures were then analyzed, as is.

DESI-MS and Sample Preparation

Samples for DESI-MS analysis were prepared by placing a small deposit (~ 1–2 mg) of the ground solid mixture onto the surface of double-sided tape (Scotch™ brand, Avon, OH) mounted onto a glass slide. The powder was pressed into the adhesive with a glass stir rod, and excess sample was removed using a stream of nitrogen gas. No significant background ions attributed to the adhesive were observed during direct analysis via DESI-MS.

Positive and negative ion spectra were recorded on a Thermo Fisher LCQ Fleet ion trap MS using an Ominspray DESI source from Prosolia, Inc. (Indianapolis, IN, USA). DESI-MS parameters utilized for this experiment were similar to those commonly reported in the literature, [28] which included a spray solvent flow rate of 3.0 μl min-1, spray voltage of 3.5 kV, a spray angle of 45°, and a nebulizing nitrogen gas pressure of 120 psi. Spectral data were collected using an average of 5 μsec/MS scan and an ionization time of 250 msec. Spray solvent compositions examined during this study were (i) water, (ii), methanol, or (iii) 1:1 methanol:water.

Spectroscopic Methods

FTIR spectra were recorded by placing the solid product on the ATR crystal of an Agilent Cary 630 FTIR spectrophotometer, collecting spectra in the 4000–400 cm−1 range. Powder X-ray diffraction (pXRD) patterns were collected at room temperature on a Bruker D8 Advance diffractometer equipped with a Vantec 2D-detector using Cu-Kα radiation in reflection mode, utilizing open sample cups for analysis. Collected data were analyzed and fitted to database patterns using the Evaluation (EVA) Application 7.001 software from SOCABIM SAS (1996–2001). Solid-state 13C-NMR was used to verify the presence of ligand exchange after grinding in the absence of any solvent, as C moieties directly attached to -SH groups are shifted significantly downfield after coordination of thiolate ion with gold, whereas there is little change in the 1H-NMR spectra. To confirm exchange reactions of target thiols with auranofin, cross-polarized magic angle spinning (CPMAS) spectra were recorded using a three-channel, 11.7 T Varian 500 MHz spectrometer operating at 125.7 MHz for 13C with a sample spinning frequency of 11.111 kHz.

Results and Discussion

Auranofin (Et3PAuStagl) and its precursor Et3PAuCl were systematically reacted via mechanochemical processing with thiols differing in net charge and molecular weight. As gold(I) compounds are known to undergo facile ligand exchange reactions in aqueous solution, it was reasonable to expect that finely divided reactants might also, under the pressure of grinding, undergo ligand exchange reactions. Utilizing DESI-MS, mass/charge ratios (m/z) of various reaction products, complemented by the unique isotopic signatures of chlorine, oxygen, sulfur, and other ligand constituents, were helpful in exact speciation of ionic species observed.

Representative DESI-MS spectra for individual gold-based reactants can be seen in Fig. 1. Collected spectra for Et3PAuCl are relatively simplistic but dependent on the spray solvent system for the desorption/ionization processes of DESI-MS. Et3PAuCl is a neutral linear molecule, wherein the chloride ligand is weakly bound to gold compared to 2,3,4,6-tetra-o-acetyl-β-1-d-glucopyranose (taglS−) in Et3PAuStagl.

Figure 1
figure 1

(a) Positive ion DESI-MS spectrum for Et3PAuCl using a methanol-only spray solvent, exhibiting a base peak at m/z 665 for [(Et3PAu)2Cl]+. (b) Isotopic distribution for m/z 665 moiety, confirming the presence of chlorine. (c) Positive ion DESI-MS spectrum for Et3PAuCl using a methanol:water (1:1) spray solvent system, showing a base peak lacking the chloride isotopic distribution at m/z 665, corresponding to [(Et3PAu)2(μ-OH2)(μ-OH)]+. (d) Representative positive ion DESI-MS spectrum for auranofin (Et3PAuStagl), showing protonated (m/z 679), sodiated (m/z 701), and an adduct (m/z 993) of the drug. (e) Representative negative ion DESI-MS spectrum for auranofin (Et3PAuStagl), showing only a gold-bound adduct of the molecule at m/z 923 and is marked by overall low intensity

Full size image

In methanol-only solvent conditions (Fig. 1a), Et3PAuCl produced a single dominant signal for [(Et3PAu)2Cl]+ (m/z = 665), exhibiting a chloride isotopic pattern (exemplified in Fig. 1B) that results from ligand exchange, and lower intensity peaks for [Et3P=OH]+, [(Et3P)2Au]+, [(Et3P=O)2Au]+ and [(Et3PAuCl)2Na]+ at m/z 135, 433, 465 and 722, respectively; alkali earth metal adducts are commonly seen in spray-based ambient MS methods, attributed to trace metals in examined samples and commercial solvents. The species distribution is markedly different in a 1:1 methanol-water solvent system (Fig. 1C), where signals were observed for [Et3PAu]+ (m/z 315), [Et3PAu(OH2)]+ (m/z 332), and [(Et3PAu)2(μ-OH2)(μ-OH)]+ (m/z 665, no Cl−isotopic signal). The Et3P=O forms by air oxidation, [(Et3P)2Au]+ by ligand scrambling, and [(Et3PAuCl)2Na]+ by ion-pairing with trace sodium. Water is a more effective ligand than methanol and is able to displace chloride to form an aqua complex, [Et3PAu(OH2)]+; this complex appears to be formed in the water-methanol system, which aggregates to form [(Et3PAu)2(μ-OH2)(μ-OH)]+ due to the presence of methanol that in turn dissociates to [Et3PAu]+ during desolvation. Representative DESI-MS spectra for auranofin can be seen in Fig. 1d (positive ion mode) and e (negative ion mode).

Data suggest that both Et3PAuStagl and Et3PAuCl undergo significant ligand exchange and ligand scrambling reactions in the solid state, represented in Scheme 1. In fact, sufficiently pure (l-cysteinato-S)-triethylphosphine gold(I) was able to be obtained from grinding accurately weighed triethylphosphine gold(I) chloride and l-cysteine in a stoichiometric ratio with a practical yield (> 98%); this offers a marked improvement over early attempts to isolate this compound with purity from solutions, which was unsuccessful due to it’s instability in the solution-phase [2, 29]. The elemental analysis (CHNS) and gold composition (by atomic absorption) of the product reasonably matched theoretical calculated values (Table 1).

Scheme 1
scheme 1

Ligand exchange reactions of auranofin using water as spray solvent in the DESI-MS experiment (top) and with l-cysteine (bottom).

Full size image
Table 1 Elemental (CHNS) analysis of (l-cysteinato-S)-triethylphosphine gold(I)
Full size table

Positive ion DESI-MS of Et3PAuStagl in the absence of added thiols (as seen in Fig. 1d) exhibited signals for protonated, sodiated and potassiated cations of the compound at m/z 679, 701, and 717 (low intensity), respectively. Additional minor signals observed at m/z = 119, 433 and 993 were assignable to [Et3PH+], [(Et3P)2Au]+ and [(Et3PAu)2(μ-Stagl)]+, respectively. The single dominant peak in the negative ion spectrum (Fig. 1e) was [Au(Stagl)2]− (m/z 923). Et3PAuStagl has been characterized via crystallographic means as a linear complex with dAu-S = 2.29 and dAu-P = 2.26 Å [29]. As the Au–S and Au–P bond lengths are close to each other, there is possibility of ligand exchange/scrambling at both sites when Et3PAuStagl is allowed to react with other soft ligands. Thus, it was expected that various species would be formed according to the scrambling equilibria accounting for these species as:

$$ 2{\mathrm{Et}}_3\mathrm{P}\mathrm{AuStagl}\rightleftharpoons {\left[{\left({\mathrm{Et}}_3\mathrm{P}\right)}_2\mathrm{Au}\right]}^{+}+\left[\mathrm{Au}{\left(\mathrm{Stagl}\right)}_2\right] $$
(1)
$$ 3{\mathrm{Et}}_3\mathrm{PAu}\mathrm{Stagl}+{\mathrm{H}}^{+}\rightleftharpoons {\left[{\left({\mathrm{Et}}_3\mathrm{PAu}\right)}_2\mathrm{Stagl}\right]}^{+}+{\left[\mathrm{Au}{\left(\mathrm{Stagl}\right)}_2\right]}^{-}+{\mathrm{Et}}_3{\mathrm{PH}}^{+} $$
(2)

These species have previously been observed in ESI-MS [14], where they exist in a stable equilibrium with the neutral Et3PAuStagl molecule, but were not observed in methanolic solution by NMR-based methods.

Mechanochemical reaction products identified via DESI-MS for either Et3PAuStagl or Et3PAuCl with a selection of individual thiols (i.e., CySH, o-MeCySH, n-AcCySH, and GSH, can be seen in Table 2, with indication of the cognizant spray solvent system included. Collision-induced dissociation (CID) MS/MS spectra of target precursor ions were used to assist in the assignment of molecular formula and deducing ligand exchange and scrambling mechanisms. Representative DESI-MS and MS/MS spectra from the mixture of auranofin (Et3PAuStagl) with l-cysteine can be seen in Fig. 2. For this combination, positive mode DESI-MS (Fig. 2a) yields characteristic products such as [(Et3PAu)2SCy]+ (m/z = 750), with negative mode exhibiting [taglSAuSCy]− (m/z = 680); additional signatures stemming from Et3PAuStagl and CySH were observed in methanol and methanol-water (1:1) spray solvents, whereas only one signal due to [(Et3PAu)2SCy]+ (m/z = 750) appeared in water. (Table 2). DESI-MS/MS spectra from the m/z 680 negative ion precursor yielded fragment ions corresponding to losses of the CH3COOH moiety at m/z 500 (n = 3) and 620 (n = 1), indicative of a cysteine-containing reaction product.

Table 2 DESI-MS data collected from examined solid-phase reactions
Full size table
Figure 2
figure 2

Representative DESI-MS mass spectrum collected from a mixture of auranofin (Et3PAuStagl) with l-cysteine (CySH) in (a) positive ion mode, showing reaction-related signatures such as [(Et3PAu)2SCy]+ at m/z 750, and (b) negative ion mode, exhibiting [taglSAuSCy]− at m/z 680. DESI-MS/MS of the m/z 680 precursor, yielding product ions corresponding to losses of the CH3COOH moiety at m/z 500 (n = 3) and 620 (n = 1)

Full size image

Formation of [Et3PH+] was also confirmed by FTIR spectra (seen in Fig. S5), where the band at 2341 cm−1 (νPH) appeared for reaction products of various thiols with Et3PAuStagl and Et3PAuCl. For Et3PAuCl reactions, only chloride exchanged with the thiol ligands, as indicated by the DESI-MS signals due to [(Et3PAu)2SCy]+ (m/z = 750). In these products, the band around 2550 cm−1 due to νSH in thiol ligands in corresponding FTIR spectra were absent, which confirmed the ligand exchange through thiol groups. Complimentary pXRD spectra, as seen in Fig. 3, also suggested the presence of phase changes stemming from the mechanochemical reaction. While spectra were not well resolved, changes in intensity and position of diffraction peaks were observed.

Figure 3
figure 3

Powder X-ray diffraction pattern of Et3PauStagl (bottom), n-AcCySH (middle), and reaction mixture of Et3PAuStagl + nAcCySH after grinding (top)

Full size image

Similarly, Et3PAuStagl was ground in equimolar amounts with o-MeCysH, N-AcCySH, and GSH and subjected to DESI-MS analysis, with representative MS and MS/MS spectra seen in Figs. S1 through S3 in the Supplementary Materials; tabulated ion assignments can also be seen in Table 2. Analogous ionic species, except [Et3PAuSR]−, were found to be present in the reaction product of Et3PAuStagl with o-MeCysH (Fig. S1). In case of GSH, the mixed dithiolato anion [taglSAuSG]− was not observed (Fig. S3). Of these thiols, N-AcCySH exhibited the slowest reactivity, as very weak signals were observed after grinding the mixture for ~ 1 h in contrast to 7–10 min in other reactions. The signals at m/z = 792 [(Et3PAu)2SNAcCy]+ (Fig. S2A) and m/z = 722 [taglSAuSnAcCy]− (Fig. S2B) were observed only when methanol was used as spray solvent. This may be due to the high pKSH (9.5) value of n-acetylcysteine, decreasing its reactivity.

For comparison purposes, all of the above experiments were repeated using physical mixtures of separately ground reactants (i.e., no mechanochemical grinding of mixed reactants). Very weak signals, some approaching detection limit (S/N = 3), for some of the ionic species listed in Table 2 could be observed, which clearly demonstrates that ligand scrambling and exchange readily takes place as a result of grinding the reactants together. This was further verified by collecting DESI-MS spectra after grinding the reactants for differing time periods, where it was found that the observed intensity of target ions increased correspondingly with reaction time, plateauing at the maxima after ~ 10 min for most thiols.

To assist in independently validating the DESI-MS results presented herein, solid-state 13C-NMR spectroscopy proved useful, using the ligand exchange reaction of auranofin with l-cysteine as a representative system. The CPMAS spectra collected from l-cysteine (only), 2,3,4,6-tetra-o-acetyl-β-1-d-glucopyranose (only), auranofin (only), and the overall reaction mixture (auranofin with l-cysteine) are shown in Fig. 4. A coordination shift of approx. 8 ppm in Cβ of l-cysteine, i.e., the carbon attached to sulfur, was observed that unambiguously verified bonding of l-cysteine with gold through the sulfur atom. No such shift was observed on simple mixing (without grinding) of the reactants. Thus, the NMR results help to support the observations from DESI-MS analysis of mechanochemical reactions.

Figure 4
figure 4

13C{1H} CPMAS spectra of l-cysteine (bottom), 2,3,4,6-tetra-o-acetyl-β-1-d-glucopyranose (bottom middle), auranofin (top middle) and the ground reaction mixture between auranofin and l-cysteine (top)

Full size image

Conclusions

The results of this study demonstrate that the DESI-MS technique can be successfully employed in monitoring the ligand exchange and scrambling reactions of coordination compounds in the solid state for gold-based therapeutic drugs. Further, the dependence of the observable reaction products on the utilized spray solvent suggests that this could be a tuneable variable, increasing the sensitivity and/or selectivity towards select targets. [30] The simplicity of mechanochemical syntheses coupled with the direct analysis capability of DESI-MS creates a rapid and impactful experimental design. In the iteration reported here, the full experiment (i.e., synthesis and analysis) was able to be accomplished in less than 20 min. This approach could prove useful towards accelerated screening of ligand targets for novel active pharmaceutical ingredients in gold-based drugs, particularly when isolation from the solution-phase is problematic. [2, 29]. With the increased availability of reduced-scale MS systems, [20, 31] performing similar investigations in the proximity of synthetic apparatus, potentially in a fume hood and/or in inert atmosphere, [32] is now conceivable.