disulfide linkages as a moDf for covalent molecular binding at surfaces has not been established previously.

AdsorpCon and chemical reacCons at metal sulfide surfaces play an important role in the bioavailability of metals and the global geochemical cycling. [1][2][3][4] The ability of iron sulfides and copper sulfides (including covellite, CuS) to catalyze the coupling of small-molecular precursors into nucleobases 5,6 has sCmulated great interest in understanding the possible role of metal sulfide surface interacCons in prebioCc chemistry. 4 7 Small CuS clusters have been found in the environment 8 and are an integral part of some metalloenzymes such as cytochrome c oxidase. 9- 11 Here, the interacCon of CuS clusters with adjacent S-containing residues to form disulfide bridges plays an essenCal role in enzyme funcCon. [9][10][11] These and other studies have moCvated interest in understanding the surface chemistry of metal sulfides, and parCcularly the interacCon of metal sulfides with thiol-containing molecules of relevance to biological chemistry.

showing that exposing plant leaves to CuO and Cu3(PO4)3 nanoparCcles improves the plant's innate defense mechanisms against root fungal diseases. [12][13][14]15 The biological pathways sCmulated by these nanoparCcles are disCnct from those produced by Cu 2+ ions in soluCon (e.g., CuSO4 soluCons), 13 moCvaCng a desire to understand the fundamental chemical properCes that govern the interacCons and transformaCons of Cu-containing nanoparCcles. 15 While the underlying physical and chemical pathways are not yet fully understood, 13 prior studies have established that the behavior of nanoparCcles in environmental and biological media is dictated by the adsorpCon of molecules from the surrounding matrix, forming a socalled "corona". 16,17 Yet, in most cases li*le is known about whether corona formaCon is controlled by chemically specific bonding interacCons, or whether it is dominated by physical forces such as electrostaCc interacCons. [16][17][18][19][20][21] These knowledge gaps arise because of the difficulty of probing adsorpCon and desorpCon processes on nanoparCcle surfaces in situ. 16 Here, we present results exploring the interacCons of copper sulfide (covellite) surfaces with glutathione, cysteine, and related molecules as model systems for understanding the interacCon of relevant biomolecules with copper-containing nanoparCcles. Copper sulfide in the form of covellite is an important model system for study because Cu-S interacCons are known to be important in Cu-containing biomolecules. [9][10][11] Furthermore, covellite forms two-dimensional sheets much like CuO and Cu3(PO4)2, but the low solubility of CuS facilitates study of molecular adsorpCon processes in aqueous media with reduced complicaCon from compeCng dissoluCon processes. Glutathione, a tripepCde, is widely distributed in animals, plants, and microorganisms [22][23][24] and is one of the most widely prevalent thiols in biology, frequently found at concentraCons of 0.1-10 mm [22][23][24][25][26][27] In the present work we use in situ ATR-FTIR measurements to probe the vibraConal features associated with adsorpCon and desorpCon from CuS nanoparCcle surfaces in aqueous media and couple these measurements with ex situ XPS studies to provide quanCtaCve measures of molecular surface coverage. A comparison of glutathione with related molecules having specific changes in chemical structure reveals that the accessible thiol group leads to irreversible adsorpCon to CuS, while the amino, carboxylate, and disulfide linkages are less effecCve. In contrast to the irreversible adsorpCon on CuS, glutathione binds only transiently to CuO and is rapidly removed by rinsing. Detailed XPS studies on covellite single crystals under condiCons maximizing surface sensiCvity coupled with density funcConal computaConal studies and addiConal experimental studies indicate that this selecCve, irreversible adsorpCon of thiols to CuS occurs by forming disulfide bridges. While disulfide bridges play criCcal roles controlling the structure and funcCon of biological systems, the possible formaCon of strong disulfide bridges in the form of a Solid-S-S-Molecule binding moCf has been predicted computaConally 28 but not established experimentally. The coupling of in situ and ex situ experimental measurements with ab ini,o theoreCcal methods provides new insight into molecular factors controlling selecCvity and irreversibility of molecule-surface adsorpCon. These insights contribute to a greater understanding of possible chemical approaches to control biological interacCons of copper-containing nanoparCcles.

Ethanol (200 proof) was purchased from Decon Labs, Inc, and all other reagents were purchased from Sigma-Aldrich. We synthesized CuS nanoparCcles following a modified procedure which has been reported on previously. 29 Briefly, 0.7 mmol copper (II) acetate was mixed with 3 mmol sulfur in 100 mL ethanol and sCrred for 10 minutes. The soluCon was then transferred to a Parr pressure vessel (Series 4760) and heated without sCrring to 160°C for 18 hours before cooling overnight. This was centrifuged at 11,000´ g for 10 minutes to remove the supernatant, before being repeated a further three Cmes with 18.2 MW•cm water. ParCcles were then dried overnight.

To compare how anion idenCty influences surface interacCons of metal chalcogenides, we also conducted control experiments using CuO nanoparCcles. CuO nanoparCcle synthesis followed previously published procedures. 16 Briefly, 1 mL of a 1 M CuCl2 soluCon was mixed with 2 mL of a 1 M LiOH soluCon, and diluted with 17 mL of deionized water. The mixture was sCrred for 10 minutes and heated in a CEM Discover microwave synthesizer at 160°C for 10 minutes.

The parCcles were then cleaned via centrifugaCon as described above for CuS. We have previously reported detailed characterizaCon of these CuO nanoparCcles. 12,16 RepresentaCve SEM images of the CuO nanoparCcles used here (Figure S1) show that they consist of thin platelets idenCcal in appearance to those we reported previously.

To prepare samples for scanning electron microscopy (SEM), dilute soluCons of nanoparCcles in isopropyl alcohol were sonicated and dropcast onto a conducCve silicon wafer heated to 90°C. Images were collected using a secondary electron detector on a Zeiss Supra55VP SEM. For powder X-ray diffracCon (XRD), nanoparCcles were packed in a zero-diffracCon plate with a 0.2 mm well (MTI CorporaCon) and analyzed with a Bruker D8 Advance powder X-ray diffractometer from 20-80° with a step size of 0.1°. X-ray photoelectron spectroscopy (XPS) data, unless otherwise specified, were collected on a Thermo K-Alpha system using nanoparCcles dropcast from an ethanol slurry onto borondoped diamond substrates (Element 6) using a take-off angle of 0 o (i.e., collecCng electrons emi*ed along the surface normal). Survey spectra were collected with a pass energy of 200 eV, 50 ms dwell Cme, and step size of 1 eV. High resoluCon spectra were collected with a 50 eV pass energy, 50 ms dwell Cme, and 0.2 eV step size. Detailed informaCon on analysis using XPS can be found in the SupporCng InformaCon. Samples for zeta potenCal analysis were prepared by sonicaCng a 50 mg/L soluCon of nanoparCcles in deionized water for 10 minutes. Zeta potenCal was measured with a Malvern Zetasizer Nano ZS, and Brunauer-Emme*-Teller (BET) surface area was measured using a Micrometrics Gemini VII surface area analyzer.

Samples for Raman spectroscopy were prepared via dropcasCng CuS nanoparCcles from isopropyl alcohol onto a conducCve silicon wafer heated to 90°C. Spectra were collected from 200 to 3200 cm -1 on a Horiba LabRAM HR EvoluCon Raman Spectrometer with a 532 nm excitaCon source and Synapse BIDD CCD detector. Spectra reported here used a 10 second acquisiCon Cme and 10 accumulaCons.

Scheme 1 shows the selecCon of molecules used for this study alongside their structure and abbreviaCons. We chose glutathione as a model system because of its widespread presence in biological Cssues. 30 Other molecules were selected to establish the role of specific molecular funcConal groups on the interacCon with CuS.

We first prepared 4 mL of a 1 mg/mL soluCon of nanoparCcles in ethanol, dispersed with sonicaCon. This was dropcast onto a germanium 10-bounce a*enuated total internal reflecCon element (Specac, 75 mm long by 3 mm thick,) before drying at 400°C for 4 minutes. Spectra were collected on a Bruker Vertex 70 Fourier-transform Infrared (FTIR) spectrometer using a Specac Gateway ATR accessory with a 550 µL flow through top plate. In these experiments, we acquired FTIR spectra conCnuously (acquisiCon Cme of approximately 4 minutes per spectrum at 4 cm -1 resoluCon) throughout 3 steps: iniCal equilibraCon with pure water (30 minutes), exposure to molecule of interest (90 minutes), and pure water rinse (90 minutes). A constant 1 mL/min flow water was maintained throughout. Absorbances at various Cme points were calculated as A=-log 10 I sample (υ &) I ref (υ &) using the last spectrum of the iniCal rinse step as the reference spectrum I ref (υ *). Control experiments using 2-mercaptoethanol were performed in a similar manner except that before rinsing, nanoparCcles were exposed to a 5% w/w soluCon of 2-mercaptoethanol for 30 min. Unless otherwise specified, all soluCons used in ATR-FTIR measurements were 65 µM concentraCon. X-ray photoelectron spectroscopy Samples were prepared by dropcasCng a nanoparCcle slurry onto boron-doped diamond, and sealing this substrate into a flow cell to mimic the condiCons used in the ATR-FTIR experiments. Using an NE-1000 syringe pump (New Era Pump Systems, Inc.) dropcast nanoparCcle films were equilibrated with water, exposed to model molecules, and rinsed following the same procedures as above for ATR-FTIR studies. XPS exposure experiments used a flow rate of 0.2 mL/min, yielding a linear flow velocity similar to that of the ATR-FTIR studies.

For the 2-mercaptoethanol exposure, the process was the same as above, except that the sample was addiConally exposed to a 5% w/w aqueous soluCon of 2-mercaptoethanol prior to rinsing. These samples were then dried for analysis. Instrumental parameters for these experiments are the same as described above. StaCsCcal analysis was conducted using JMP, Version 17.2.0 (SAS InsCtute Inc., Cary, NC).

The sampling depth of XPS is controlled by the inelasCc mean free path (typically ~2 nm) of the electrons that are emi*ed and the angle at which electrons are collected. CollecCng electrons emi*ed at very shallow angles with respect to the surface plane yields improved sensiCvity to the very outermost region of the sample, improving the ability to probe the surface chemistry of importance to our studies. This approach requires samples with exposed faces that are flat and several mm in size so that the electron emission angles are well-defined with respect to the surface plane. Since covellite samples meeCng these requirements are not readily synthesized in the laboratory, we used natural single-crystal covellite samples (Excalibur Mineral CorporaCon). These were cleaned with 12 M HCl prior to use. Covellite was exposed to 65 µM glutathione on an orbital shaker for one hour before rinsing. XPS measurements using take-off angles of 80 o from the surface normal were conducted using a Phi VersaProbe III system. To achieve a well-defined take-off angle, we rotated the sample appropriately and introduced an aperture between the sample and the analyzer. High-resoluCon spectra were collected with a 55 eV pass energy, 50 ms dwell Cme, and 0.1 eV step size. Details of energy calibraCon and peak-firng are presented in the SupporCng InformaCon.

To determine the impact of molecular adsorpCon on the release of copper from CuS, we carried out dissoluCon studies in triplicate using two soluCon sets: a 'plant-like' medium containing 890 µM malic acid and 1.71 mM citric acid, and a second medium containing those with the addiCon of 1 mM glutathione. SoluCons were removed ater 2 hours, 8 hours, 24 hours, and 4 days, before being centrifuged at 13,100 g (Eppendorf MiniSpin Plus) for 10 minutes, and filtered through 0.1 µm syringe filters. SoluCons were then acidified with 2.5%

HNO3 and analyzed by ICP-MS (8900 Triple Quadrupole, Agilent) with y*rium as an internal standard.

To understand the structure of CuS surfaces and the thermodynamics associated with adsorpCon of cysteine as a model thiol, we used ab ini,o atomisCc thermodynamics calculaCons. Cysteine was chosen over glutathione to reduce the number of possible molecular conformaCons. A detailed explanaCon of the underlying methodology is contained in the SupporCng InformaCon. Briefly, periodic density-funcConal theory (DFT) calculaCons were performed on model CuS structures using the generalized gradient approximaCon of Perdue, Burke, and Brinkerhoff (PBE) as implemented within the DMol 3 code. 31,32 All-electron calculaCons were done using a double numeric basis set with polarizaCon (DNP). 33 All structures were subject to full geometry opCmizaCons, where self-consistent field energies were converged to at least at least 3 × 10 -6 eV, and the opCmizaCon reached a threshold of at least 3 × 10 -5 eV or unCl the force on each atom converged to < 1 meV/Å. Molecular calculaCons were also carried out with the use of periodic boundary condiCons, with O2 and H2 molecules described in the gas phase. H2O and cysteine molecules were modeled with the use of the implicit solvaCon model COSMO, wherein the calculated gradients include forces between the solute and screening charges to simulate the electrostaCc effects an aqueous environment with a dielectric constant (𝜀) of 78.54. 34,35 The bulk covellite CuS structure, space group P63/mmc (194) was modeled in a hexagonal unit cell using a Monkhost-Pack 36 8×8×4 k-grid, yielding opCmized larce constants of a=b=3.838 Å and c=16.583 Å, in reasonable agreement with experimental values (a=b=3.794 Å and c=16.341 Å) and previously reported theoreCcal values. [37][38][39][40] The k-grid is folded down to 6×6×1 for the opCmizaCon of bare and hydrated surface slabs.

As shown in Figure 1, sulfur has two chemically disCnct sites, comprised of Cu-S-Cu ("sulfide") and Cu-S-S-Cu layers ("disulfide") layers. Prior studies of single-crystal covellite 41 , nanocrystalline covellite prepared through different synthesis routes, [42][43][44] and computaConal studies of covellite thermodynamics 37,40 have uniformly concluded that the (001) plane has the lowest surface energy and is the predominant face exposed. SEM images of our nanoparCcles (vide infra) show thin platelets having almost perfect hexagonal symmetry. We therefore focused our studies on the (001) crystal face and the different possible surface terminaCons.

The opCmized bulk CuS was cleaved along the (001) direcCon to generate the 1×1 surface models, where the different bond cleavage planes are idenCfied in Figure 1. There are five possible surfaces that can be formed that exhibit different terminaCng atom(s): two Sterminated (SS and S Pucker), two Cu-terminated (Cu T and Cu Pucker), and one CuSterminated (Flat CuS). All surface slab models were set up with inversion symmetry; in doing so, any polarizaCon effects caused by the cleaving process are negated. Slab thicknesses ranged between 13-20 Å, with at least 15 Å of vacuum between periodic slabs to avoid through-space interacCon of periodic repeats.

Figure 2 shows characterizaCon of the CuS nanoparCcles using SEM (Fig. 2a), XRD (2b), and XPS (2c). The SEM image in Fig. 2a image shows that the nanoparCcles consist of thin platelets with hexagonal habit, with outgrowths distributed across some parCcle. The anisotropic habit and hexagonal shape indicate that the broad faces expose the (001) plane, in agreement with previous experimental and computaConal studies. [41][42][43][44] Detailed analysis of SEM images using ImageJ sotware show that the average nanoparCcle size, measured by the distance between opposite edges of the hexagons, is 500±130 nm (N=258) with a thickness of 70±20 nm (N=167). Fig. 2b shows powder X-ray diffracCon data from our CuS nanoparCcles, along with a reference spectrum for the CuS covellite structure. 45 Our experimental data are in excellent agreement with the reference spectrum with no traces of other structures, thereby indicaCng a high degree of structural purity. Figure S2 includes a comparison showing that the XRD features match that of a reference covellite spectrum and are disCnct from other copper sulfide phases. Fig. 2c shows XPS survey spectra of the nanoparCcle surfaces.

Analysis of fracConal surface composiCon (details in SI) yields a Cu:S raCo of 1.08:1 and is consistent with that of the 1:1 stoichiometric raCo for bulk covellite. We addiConally performed more detailed analysis of the nanomaterial surfaces using high-resoluCon XPS and by Auger parameter analysis. Detailed analysis of the S(2p) region by XPS shown in Figure S3 reveals three set of peaks that are a*ributed to S atoms in the Cu-S-Cu layers (161.2 eV) and the Cu-S-S-Cu layers (162.0 eV), along with a third feature at higher binding energy (163.0 eV). This feature can likely be a*ributed to a combinaCon of shake-up features and chemically disCnct surface species. [46][47][48][49][50] Previous studies have shown that the modified Auger parameter, the difference in energy between Auger and photoelectron peaks for specific elements, is disCnct for various copper compounds. 51 We performed a modified Auger parameter analysis by summing the binding energy of the Cu(2p3/2) photoelectron and kineCc energy of the L3M45M45 peak from the XPS survey in Fig. 2c, and found a value of 1850.5 eV. This value is within the accepted range of modified Auger parameter values for CuS of 1850.3±0.2 eV for covellite and is significantly greater than the value anCcipated for other stoichiometric and non-stoichiometric copper sulfides. [51][52][53] We further characterized our CuS nanoparCcles using Raman spectroscopy. Figure S4 shows that the Raman spectrum is in close agreement with that expected for covellite and is disCnct from other copper sulfides. Zeta potenCal measurements show a zeta potenCal of -27±1 mV, and the BET surface area of these nanoparCcles is 11.0 m 2 /g.

Figure 3 shows data probing the interacCon of glutathione with CuS and CuO surfaces using XPS and FTIR. We used N(1s) emission intensity to quanCfy the molecular coverage of glutathione on these surfaces, using the Cu(2p) emission as an internal standard. From the areas of the N(1s) (Figure S5) and Cu(2p) regions, in conjuncCon with the inelasCc mean free paths and the atomic sensiCvity factors, we esCmated the absolute surface coverage as shown in Fig. 3a. Details of this analysis are described in the SupporCng InformaCon. Fig. 3b shows XPS survey spectra of a CuS sample that was equilibrated with pure water, exposed to glutathione, and then rinsed for 90 minutes as described above. In this experiment, the CuS nanoparCcles were dropcast onto a conducCve boron-doped diamond substrate. The XPS data show that glutathione adsorbs strongly to CuS surfaces, yielding molecular coverages of ~2 molecules/nm 2 . In contrast, interacCons with CuO shows much lower coverage, only ~0.5 molecules/nm 2 . StaCsCcal analysis shows that the difference in adsorpCon on CuS and CuO surfaces is significant at the 95% confidence interval. While the extent of adsorpCon can be impacted by many factors, zeta potenCals of CuS (-27±1 mV) and CuO (-19.7±0.3 mV) are similar. Thus, we conclude that the difference in binding affinity of glutathione on CuS vs. CuO is likely not a result of electrostaCc interacCons but is more likely due to a chemically specific interacCon.

While XPS provides informaCon about the total extent of adsorpCon, we used in situ FTIR spectroscopy to probe the changes in vibraConal spectra during exposure to glutathione and during the subsequent rinse. As described above, each in situ experiment generates spectra at mulCple Cme points, yielding both spectroscopic and kineCc informaCon. In the SupporCng InformaCon Fig. S6, we show a complete set through a typical 90-minute exposure and 90minute rinse. The in-situ spectra show vibraConal features due to interacCon of glutathione with the surfaces, plus addiConal features arising from atmospheric water vapor, atmospheric CO2, and imperfectly compensated liquid water. ExaminaCon of the compete set shows that changes in the vibraConal spectra during the exposure step and during the rinse step are each complete within tens of minutes. We therefore focus a*enCon on the last spectrum obtained while flowing glutathione ("flow") and the last spectrum of the subsequent 90-minute rinse ("rinse") for more detailed analysis, shown in Fig. 3c. These spectra are limited to the amide vibraConal region and have been background-subtracted by removing broadly sloping backgrounds. Fig. 3c also shows an infrared spectrum of glutathione measured using the same ATR crystal, but without the CuS nanoparCcles present. We note that to obtain a spectrum with comparable absorbance, it was necessary to go to concentraCons approximately 1000-fold higher (100 mM) compared to the spectra obtained with the CuS nanoparCcles (65 micromolar).

We therefore conclude that the vibraConal features observed in the in-situ spectrum at 65 mM concentraCon arise almost exclusively from glutathione adsorbed to the CuS nanoparCcles surfaces, with only negligible contribuCon from glutathione in the aqueous phase.

To glean more informaCon the glutathione-CuS interacCon, we examined the relaCve intensity of the peaks associated with glutathione in soluCon and bound to the surface. The absorbances shown in these spectra match well with those for glutathione, including a carboxylic acid C=O stretch at 1725 cm -1 , peaks associated with amides at 1636 cm -1 and 1540 cm -1 , and further complex amide bands between 1500-1100 cm -1 . 54,55 We were not able to observe the S-H vibraConal modes of glutathione either in the presence of or absence of CuS nanoparCcles; this result is consistent with prior reports that thiol group vibraCons are frequently very weak and broad due to hydrogen bonding. 56 Our FTIR spectra are consistent with the XPS data, showing that adsorpCon of glutathione occurs quicky (within tens of minutes) and irreversibly, as there is no significant decrease in amide band intensity when the 65 mM glutathione is replace with deionized water for 90 minutes.

As a comparison, we also performed similar ATR-FTIR samples using CuO nanoparCcles instead of CuS nanoparCcles. Fig. 3 shows that with CuO nanoparCcles, the FTIR spectra obtained while flowing GSH show significant FTIR intensity, but the features disappear ater the GSH is replaced with deionized water. During flow, the amide band intensity is again much larger than what would be expected from the soluCon-phase GSH alone. Therefore, we conclude that with CuO surfaces, glutathione binds weakly to the surface and is concentrated on the surface while GSH is flowing, but the surface-bound molecules are easily removed when the GSHcontaining soluCon is replaced with pure water.

From the above experiments, we conclude that glutathione binds to CuS surfaces within several minutes of exposure and that this binding is irreversible on the ~90-minute Cme scale invesCgated here. Glutathione also interacts with CuO when excess GSH is present in the aqueous phase but desorbs rapidly during the water rinse. These observaCons are consistent with the XPS results but provide addiConal insight into the adsorpCon and desorpCon processes.

The above results show that glutathione interacts strongly with CuS but not with CuO. To determine whether this interacCon is specific to the presence of thiol groups or not, we performed similar experiments using a set of model molecules as depicted in Scheme 1. As a control for the adsorpCon of glutathione, we used ophthalmic acid (OTA), an analogous tripepCde in which the thiol group is replaced with a methyl group. We also performed experiments with the small amino acids cysteine (CYS) and homoalanine (HAL) to probe the influence of molecular size. Finally, we used glutathione disulfide (GSSG) to probe the importance of free thiol groups. We characterized the molecular interacCons using XPS and FTIR in a manner similar to the experiments using glutathione.

Figure 4 shows the XPS surface coverages of the molecules described above. We again use N(1s) emission to evaluate the surface concentraCon of adsorbed molecules. The surfaces exposed to glutathione and those exposed to cysteine show significant surface N concentraCons. In contrast, surfaces exposed to homoalanine, glutathione disulfide, and ophthalmic acid exhibit N concentraCons that are significantly lower from glutathione and cysteine concentraCons at the 95% confidence interval. QuanCfying the coverage using the Cu(2p) emission as an internal standard shows that molecules with accessible thiol groups (glutathione and cysteine) yield high surface coverages, while molecules with similar structures but lacking the thiol group (i.e, homoalanine, glutathione disulfide, and ophthalmic acid) yield much lower surface coverages.

Our data suggests that surface binding of thiol-containing molecules occurs via formaCon of surface disulfide linkages. Disulfide bridges are ubiquitous in biology and are present in one-third of eukaryoCc proteins. 57 In biological chemistry, exposure to 2mercaptoethanol, which cleaves disulfide linkages, is widely used as a test for the presence of disulfide linkages. To probe this, we performed an experiment in which a CuS nanoparCcle film was exposed to glutathione as in other experiments, and then the sample was rinsed with a soluCon containing 2-mercaptoethanol (2ME). The XPS data show that exposure of the GSHexposed surface to 2-mercaptoethanol significantly decreases the surface N coverage, implying that 2ME cleaves most (but not all) of the surface-molecule bonds.

Figure 5a shows FTIR spectrum for CuS nanoparCcle films exposed to each of these molecules. In each case, the in-situ experiments show rapid (tens of minutes) adsorpCon, with li*le or no decrease in absorbance when the flowing soluCons are replaced with pure water rise. The spectra for ophthalmic acid, glutathione disulfide, and homoalanine have been magnified 5-fold due to the low intensity of peaks associated with adsorpCon for each of these molecules. The spectra in Fig. 5a fully corroborate the XPS data for each of these molecules,

showing that cysteine and glutathione, both of which have accessible thiol groups, adsorb strongly, while ophthalmic acid, glutathione disulfide, and homoalanine do not. Figure 5b shows in situ FTIR data for of a sample that was exposed to GSH, and then the same sample ater flowing 2-mercaptoethanol (2ME) over it. In agreement with the XPS data above, the FTIR data show that exposing the surface-adsorbed GSH molecules to 2ME cleaves the GSH molecules from the surface.

To confirm that the chemical selecCvity inferred from the above data, we performed two addiConal sets of control experiments. Measurements of the IR spectra of the different molecules using a single-bounce Ge ATR element in the absence of any nanoparCcles (Figure S7, using 100 mM concentraCon) show that the molecules in soluCon all have IR absorpCon features with similar intensiCes. This indicates that the molar absorpCviCes, and therefore the analyCcal sensiCvity to each molecule in soluCon, are comparable to one another. The data in Figure S7 were obtained at mich higher concentraCon (100 mM) compared to those used elsewhere in this work (65 µM). Figure S8 shows IR spectra obtained on the mulCple-bounce Ge ATR element, again in the absence of nanoparCcles, at the 65 µM concentraCon used in other experiments reported here. These spectra show much lower absorbances compared to those in Figure 5; this large difference indicates the contribuCons to the spectra in Fig. 3 and Fig. 5 from free molecules in soluCon are negligible relaCve to those adsorbed on the surface.

The presence of internal disulfides as an integral part of the covellite crystal larce complicates detecCon of possible disulfides forming at the CuS-molecule interface. In an effort to isolate the spectroscopic features of the glutathione-CuS interfacial bonds, we used highangle XPS to maximize the surface sensiCvity and therefore specificity of our measurements. If a planar sample is Clted such that only electrons emi*ed close to the surface plane can reach the analyzer, electrons arising from the bulk are preferenCally sca*ered and reduced in intensity compared to those from the outmost surface. CollecCng electrons emi*ed at 80° from the surface normal (compared to 0 o used elsewhere in this work) yields an improvement of ~6x in sensiCvity to the surface compared to the near-surface bulk region. surface normal for the same covellite single-crystal sample, both before and ater exposure to glutathione. Peak-firng using procedures idenCcal to those used in Figure S3 yielded 2p3/2 peaks at 161.2,161.9 eV, and 163.2 eV, along with their respecCve 2p1/2 components shited by 1.18 eV. These energies are nearly idenCcal to those in Figure S3. Peak-firng for these components resulted in a reduced χ 2 =1.12, indicaCng that the residual error between data and fit is only slightly greater than that a*ributable to the noise.

The glutathione-exposed sample spectrum shows an increase in the 2p3/2 feature at higher binding energy shiting slightly to 163.4 eV. As noted earlier, broad emission in this range on the starCng surface can arise from shake-up features and from surface species. In order to isolate the glutathione-induced changes without addiConal peak firng, we determined the S(2p) changes upon molecular adsorpCon by direct subtracCon of the starCng surface from the glutathione-exposed sample. This difference spectrum shows two disCnct peaks, at 163.4 eV and 164.6 eV, with a 2:1 intensity raCo. The 1.2 eV splirng and 2:1 intensity raCo are both consistent with these peaks arising from the 2p3/2 and 2p1/2 spin-orbit of a single chemical form of sulfur. While the precise structure of the exposed CuS surface remains unknown, a qualitaCve analysis of the S(2p) region provides some insight. Compared with neutral S atoms, S atoms in bulk CuS are in a more ionic environment; the electronegaCve S atoms withdraw electron density from Cu, leading to parCal negaCve charges on the S atoms and core electrons that are more weakly bound. Thus, the 2p3/2 binding energy of S atoms bound to nearby Cu is 161.2 eV, compared with approximately 163 eV for polysulfides with S atoms in a nearly neutral environment [58][59][60] . The new feature we observe ater glutathione exposure has a binding energy close to that of the neutral polysulfides, suggesCng that that S atom of glutathione is binding to another S atom of the surface rather than an exposed Cu 2+ ion.

Our XPS and FTIR data suggest that glutathione, cysteine, and presumably other thiols also bind to CuS surface via formaCon of surface disulfide linkages. We performed density funcConal theory calculaCons to assess whether this type of bonding moCf is feasible thermodynamically.

To explore the overall thermodynamic stability of binding via surface dithiol linkages vs.

other possible bonding configuraCons, we used DFT calculaCons of CuS surfaces both in the presence and absence of cysteine as a model adsorbate. These calculaCons included the influence of water on the surface terminaCon and the dielectric constant of the medium. Figure 7 compares the different funcConal groups that can form on the CuS covellite (001) surface when hydrated. We expect that surface sites terminated with undercoordinated S atoms will protonate to form thiol (SH) funcConal groups and that surface sites exposing undercoordinated Cu atoms will hydroxylate to form copper hydroxyl (CuOH) funcConal groups.

Cleavage along a (001 (2)

Previous work characterizing aqueous metal oxide minerals has demonstrated the tendency for surface funcConalizaCon to influence the thermodynamic stability of different facets and terminaCon schemes. [61][62][63][64] The values of ∆Gtotal for the formaCon of SH and CuOH funcConal groups are presented below in Table 1. A posiCve ∆Gtotal value indicates that the formaCon of hydra0ng condi0ons. Atoms are represented using ball-and-s0ck models, where

Using the previously determined stable SH terminated structures (SS and S Pucker), 2×2

supercells were generated to model adsorpCon at reasonable coverage. Cysteine adsorpCon was modeled as an analog to glutathione to reduce the computaConal cost and limit the number and types of surface funcConal groups that could possibly engage the CuS surface, and to test the thermodynamic likelihood of S-S bond formaCon at the CuS surface. Inner-sphere adsorpCon on the OH-terminated Cu T surface was deemed unfavorable as no stable adsorpCon geometry could be idenCfied without the surface and/or adsorbate undergoing significant structural changes that cannot be directly compared to the SH-terminated surfaces.

Details of the mechanisms modeled here can be found in the SupporCng InformaCon.

The opCmized adsorpCon geometries are shown in Figure 8 for the three possible structures; the Eads values for the formaCon of each complex is indicated below. We note that the Eads values for the formaCon of all three complexes are all favorable and very similar, ranging only by 0.05 eV regardless of if the surface terminates with SS or S Pucker. Comparing the two structures that bind cysteine to the surface through S-S bonds (labeled with (1)), both adsorpCon complexes have similar C-Scys bond lengths to the opCmized isolated cysteine molecule, but the S Pucker (1) structure has slightly smaller Scys-Ssurf bonds compared to the SS The number in parentheses indicates which reac0on scheme was used to compute Eads. Atoms are represented using ball-and-s0ck models, where the Cu, S, O, H, C, and N atoms are colored orange, yellow, red, white, gray, and blue, respec0vely.

(1) complex. This slight difference in the one bond length results in a smaller Ssurf-Scys-C bond angle of 104.4° for the S Pucker (1) adsorpCon geometry compared to the SS (1) angle of 110.6°.

Considering the SS (2) structure, the Ssurf-2-Ssurf-1 bond of SS ( 2) is more like the Ssurf-Scys bonds of SS (1) and Pucker (1) than the Ssurf-1-Scys bond. The formaCon of S-S-S bonds causes lengthening of the Scys-Ssurf-1 bond compared to the Scys-Ssurf bonds compared to the (1)

structures. The C-Scys bond in SS ( 2) is also slightly longer but seems to be less sensiCve to changes in local environment than the Scys-Ssurf bonds. The Ssurf-1-Scys-C bond angle is 103.6°, which is smaller than the (1) analogs, due to the short Ssurf-1-Scys bond length. It is also worth noCng that the formaCon of a trisulfide bond, shown as SS (2), is just as favorable as formaCon of the disulfides in S Pucker (1) and SS (1). Given that sulfur is known to form polysulfides favorably, 65 this is not an enCrely unexpected result, although the similar Eads values for each interacCon in the CuS-cysteine system means that experimental verificaCon is difficult.

DFT calculaCons were also performed to assess the vibraConal frequencies of anCcipated surface species. The results, shown in SupporCng InformaCon, indicate that the vibraConal modes of the S-S surface species are predicted to be very close to those of the bulk sulfur-related Cu-S-S-Cu modes. This is consistent with experimental efforts that we made to detect surface species by Raman spectroscopy of the GSH-exposed surface, which were unsuccessful due to the large background from the bulk S-S modes.

Previous studies of nanoparCcles in aqueous media and biological systems have shown that the composiCon of the aqueous matrix can significantly impact the rates of dissoluCon and other transformaCons. Small organic acids, especially citric acid, have been shown to play an important role [66][67][68] and are naturally present in the at millimolar concentraCons in many plant species [69][70][71] and have been used in simulated xylem media for studies on agriculturally relevant nanomaterials 12,72 . In order to idenCfy how copper release is affected by the adsorpCon of glutathione, we invesCgated dissoluCon of Cu into two different 'plant-like' aqueous media: one containing 890 µM malic acid and 1.71 mM citric acid at pH 6, and one containing both of those alongside 1 mM glutathione.

Figure 9 shows ICP-MS Cu data for nanoparCcles exposed to the aforemenConed sample media for duraCons of 2, 8, 24, and 96 hours, alongside a media blank containing no nanoparCcles (shown at Cmepoint 0). The parCcles in typical 'plant-like' media release a small concentraCon of Cu 2+ , yielding concentraCons of approximately 1.3 µM Cu 2+ ater 96 hours.

InteresCngly, however, nanoparCcles exposed to the media containing glutathione results in very li*le Cu release (0.1 µM). Thus, we conclude that the high coverage of glutathione effecCvely thwarts dissoluCon and release of Cu 2+ , even in the presence of chelaCng weak acids.

The work reported here shows that thiol groups of small molecules of biological significance bind irreversibly to covellite surfaces, while other common funcConal groups (e.g., carboxylate groups and amino groups) induce much weaker interacCons. This conclusion is consistent with previous reports that that cysteine-containing pepCdes have parCcularly strong affiniCes for binding to ZnS. 73 . Our experimental and computaConal results both point to the formaCon of molecule-surface disulfide linkages in which the free thiol group of the molecule binds to sulfur atom of the CuS larce as depicted in Fig. 8 as the surface binding moCf. This conclusion is consistent with a prior molecular dynamics study of amino acid interacCons with ZnS. 28 That study 28 calculated that cysteine formed a covalent bond to ZnS via a disulfide linkage similar to that we report here with a binding energy of 98 kJ/mole, while other amino acids had much weaker interacCons <4.3 kJ/mole. Our results on CuS are consistent with this, with overall adsorpCon energies of ~ -1.5 eV or approximately 140 kJ/mole for cysteine on CuS (001). Prior studies of thiols binding to gold surfaces have also reported formaCon of disulfides in a different moCf, in which two adjacent surface-bound alkanethiols join to form a S-S bond; 74 however, the disulfide formed on gold has both S atoms bound to the surface and is therefore structurally and chemically disCnct from that we describe here.

Our studies show that while of CuS is relaCvely slow, adsorpCon of glutathione further reduces this process, even in matrices with relaCvely high concentraCons of small acids that are able to chelate Cu 2+ . These dissoluCon studies further support the favorable formaCon of disulfide bond formaCon in matrices more complex than deionized water. While most environmentally relevant matrices are highly complex, our work suggests that the presence of even low concentraCons of glutathione and related thiols may be sufficient to control the formaCon of molecular coronas in more complex environments, such as in the xylem or phloem of plants or in other environmentally relevant matrices. This work also suggests that CuS has the potenCal to sequester glutathione, cysteine, and other thiol-containing molecules from complex environments; the associated reducCon in concentraCon of these thiol-containing molecules represents another potenCal pathway for biological impact of CuS nanoparCcles.

Glutathione contributes to a number of criCcal funcCons in plants, including stress signaling and communicaCon 22,75 , removal of reacCve oxygen species (ROS) 76 , and the chelaCon of toxic heavy metals 77 .

While this study was moCvated by a desire to understand the fundamental surface chemistry associated with Cu-containing nanoparCcles underlying their biological impact, the chemically specific formaCon of molecule-surface bonds has potenCal impact in other areas.

Our studies show that while the disulfide bridges are stable under ordinary condiCons even in the presence of small organic acids, the molecule-surface bond can be cleaved through introducCon of 2-mercaptoethanol. The ability to form chemically specific molecular layers than can be selecCvely cleaved under mild condiCons through the use of specific chemical reacCons (e.g., reducCon of the disulfide many possible applicaCons in more advanced chemical manipulaCon of CuS surfaces.

used in this study had a zeta-potential of mV and a BET surface area of 13.2 m 2 /g. Figure S1 shows representative SEM of these nanoparticles.

We conducted XPS analysis using Shirley backgrounds and Voigt functions in CasaXPS software. 1 We obtained the composition of the near-surface region using Cu:S=

A S S S λ S,CuS where Ax is the area of the relevant XPS peak, SX refers to the atomic sensitivity factor for each element, and λx, CuS is the inelastic mean free path of electrons of the respective XPS features.

The sensitivity factors used here for surface quantification are SFN(1s)=1.68, SFC(1s)=1, SFS(2p)=1.88, and SFCu(2p3/2)=18.15. These sensitivity factors are specific to the Thermo K-Alpha instrument used for analysis. Inelastic mean free paths used here are for electrons from the element of interest through the CuS substrate, λN(1s)=1.968 nm, λS(2p)=2.379 nm, and λCu(2p3/2)=1.180 nm. Quantification of nitrogen surface coverage was calculated using the following equation: Figure S4 shows Raman CuS nanoparticles compared to spectra for other stoichiometric and non-stoichiometric copper sulfides. 6 Peaks associated with covellite are present in the spectrum for CuS nanoparticles, while chalcocite and digenite are absent.

Figure S5 shows compared high resolution N(1s) spectra for CuS nanoparticles exposed to GSH, CYS, OTA, HAL, GSSG, and GSH followed by a disulfide reducing agent. These data have been divided by the relative area of the Cu(2p3/2) peak used for determining their surface coverage, in order to account for fluctuations in instrument data collection, spots on each sample, and the emission intensity of the substrate. glutathione on the surface of CuS over the course of these experiments. As described earlier, these data are background-corrected relative to the end of the pure water equilibration step. Each spectrum takes approximately 4 minutes to collect. The spectra with solid lines are during the adsorption flow step, with peaks growing in that are representative of glutathione in solution. After rinsing, shown with dashed lines, the peaks and their relative position and intensity have not significantly changed. During rinse GSH exposure Absorbance

Figure S7 shows ATR-FTIR of these biomolecules in water at a concentration of 100 mM. These spectra show that all biomolecules used here absorb similar amounts of infrared light to each other and provide structures for comparison to other ATR data analyzed here.

Figure S8 shows infrared spectra of each biomolecule after being flowed over a germanium ATR element for 90 minutes, followed by 90 minutes of rinsing. initio Thermodynamics: Surface Stability

Ab initio atomistic thermodynamics 7-10 was used to predict surface free energies. Surface free energies in vacuum were calculated as a function of the chemical potential:

where G refers to the Gibbs free energy of the relaxed slab model, 2A refers to the surface area of the two equivalent slabs, µi(T,p) is the chemical potential of species i, and Ni is the number of atoms of species i. Values of Gibbs free energy were obtained by incorporating the vibrational zero point energy in the DFT total energy; entropic contributions to the total energy are negligible under the presumed conditions. The surface free energy approach of Reuter and Scheffler is well-documented in other work and is summarized for the CuS system of interest here. The expression for γ(T,p) can be rewritten as: 𝛾(𝑇, 𝑝) = ! [𝐺 %&'( )*+ -𝑁 )* 𝜇 )* (𝑇, 𝑝) -𝑁 + 𝜇 + (𝑇, 𝑝)] (S2)

where 𝐺 %&'( )*+ is the Gibbs free energy of the CuS surface slab with two equivalent surfaces. The free energy of the bulk CuS (𝐺 (*&, )*+ ) can be used to rewrite Equation S2 so that it is a function of µS:

It is important to consider the thermodynamically accessible range for the sulfur chemical potential. The S-poor limit can be defined as the chemical potential from bulk CuS, while the Srich limit is defined as the chemical potential from bulk α-S, as done in previous work. 11 We assessed the surface free energy at 0 K to determine the surface stability under Under S-rich conditions, the trend in increasing surface stability is: Cu T < S Pucker < Cu Pucker < SS < Flat CuS. As the conditions change to the S-poor regime, the SS surface becomes increasingly less stable, while the Cu T surface becomes slightly more stable. The trends in surface free energy agree well with previous findings on the bare CuS surface terminations, with the Flat CuS being the lowest energy structure and the Cu T being the highest. 11,12 DFT + Thermodynamics: Surface Functional Groups Under aqueous conditions undercoordinated S and Cu atoms will react with nearby species such as water and water products to regain coordinative saturation. Calculations assumed respect to µ s , the chemical poten1al of S. The leZhand side of the plot corresponds to the Spoor environment and the right corresponds to the S-rich environment.

that bulk-terminated surfaces exposing Cu atoms were hydroxylated to form CuOH surface functional groups and surfaces exposing under-coordinated S atoms were protonated to form SH functional groups.

The thermodynamics of surface functional group formation were assessed using the DFT solvent-ion approach, which approximates the free energy change associated with the dissolution or deposition of surface atoms and the adsorption or desorption of HaOb species (H, OH, O, and H2O). The method requires a reference surface structure and the partitioning of the overall process into elementary steps. The contribution of each step to the overall change in Gibbs free energy is determined either using DFT total energies or tabulated experimental data regarding the aqueous electrochemical properties and with analytical corrections to account for changes from standard conditions. 13 This approach has been applied to a variety of systems and applications-from predicting the metal release or dissolution from battery cathode materials [14][15][16][17] to understanding the sorption preferences of contaminants at the mineral-water interface. 18,19 This method has also been applied toward understanding how nanomaterials can transform under aqueous conditions. 20,21 The utility of this model allows the overall modeled reaction to be partitioned into purely-DFT components (∆G1) and components determined using experimental information (∆G2). For example, consider the formation of SH functional groups on the S-terminated surface, where the overall reaction is shown in Equation (S4):

CuS-Cu + H2O(l) à CuS-CuOH + H + (aq) (S4)

This reaction is broken down into elementary steps that can be descried either using DFT or tabulated experimental information. ∆G1 is calculated using DFT total energies and Hess's Law according to Equation (S5), where the reactants are referenced to their standard states (H2 (g) and There are two possible reaction mechanisms that were considered, shown in Figure S10.

The first reaction (Scheme 1) is applied to both the S Pucker and SS structures and involves a ligand exchange reaction with a surface SH group and the SH group of cysteine, producing a SH2 molecule. The second reaction (Scheme 2) applies only to the SS structure and assumes that the product is a H2 molecule and that S-S-S bonds bind the cysteine to the CuS surface. Adsorption energies (Eads) are calculated using Hess's Law for each of the schemes in Figure S10, where the products are the adsorption complex and a SH2/H2 gas molecule and the reactants are the hydrated surface and the cysteine molecule.

The surface free energies of the S-terminated structures that form SH functional groups were also calculated are compared in Figure S11. The chemical potential of H is referenced to H2 (g). This approach compares well to the DFT + Thermodynamics approach that references the