Tailoring NIR-II photoluminescence of single thiolated Au25 nanoclusters by selective binding to proteins

Atomically precise gold nanoclusters (Au NCs) are a fascinating class of nanomaterials that exhibit molecule-like properties and have outstanding photoluminescence (PL), which is highly dependent on their structure and chemical environment. Their ultrasmall size, molecular chemistry, and biocompatibility make them extremely appealing for selective biomolecule labeling in investigations of biological mechanisms at the cellular and anatomical levels. In this work, we report a simple route to incorporating a preformed Au25 nanocluster into a model bovine serum albumin (BSA) protein. A new approach combining small-angle X-ray scattering and molecular modeling provides a clear localization of a single Au25 within the protein to a cysteine residue on the gold nanocluster surface. Attaching Au 25 to BSA strikingly modies the PL properties with enhancement and a redshift in the second near-infrared window (NIR-II). An extensive study based on a bottom-up approach that uses mixed-ligand nanoclusters Au 25 pMBA (18−x) Cys x with x=2, 5, 18 supported by experimental data (steady state, time-resolved spectroscopy) and theoretical calculations (DFT) provides new hints at the origin of NIR-II emission in such nanoclusters and their subsequent enhancement when selectively binding to a cysteine-rich protein. This study paves the way to controlling the design of selectively sensitive probes in biomolecules through a ligand-based strategy to enable the optical detection of biomolecules in a cellular environment by live imaging. pMBA - 4-mercaptobenzoic acid) into bovine serum albumin (BSA) protein selectively anchored to one cysteine residue without drastic modication of the protein structure. The attachment of a single thiolated Au 25 nanocluster to the protein had a signicant effect on its NIR-II emission band with a 4.7-fold enhancement of the PL QY. Such water-soluble Au 25 NCs attached to BSA display bright infrared emission between 700 nm and 1300 nm, in contrast to previously reported uorescent BSA-directed Au NCs that only emit below 1000 nm. Mass spectrometry analysis revealed that the attachment of Au 25 NCs to BSA involves ligand exchange between the pMBA ligands and BSA cysteine (Cys) residues. Small-angle X-ray scattering (SAXS) provided an unambiguous localization of a Au 25 NC on the protein via a free cysteine residue, Cys34. To better understand the PL properties of Au 25 nanoclusters attached to BSA, we synthesized mixed liganded nanoclusters, Au 25 pMBA (18-x) Cys x with x=2, 5, 18, through ligand exchange. The origin of NIR-II emission in nanoclusters with mixed ligands determined by optical measurements and density functional theory reveals the key role of the metallic-ligand interface in photophysical pathways. These results pave the way for the tailored synthesis and protein bioconjugation of nanoclusters to extend their uses for diagnostic and therapeutic applications.


Introduction
Gold nanoclusters (Au NCs) are at the forefront of nanomaterials that can be used for bio-applications due to their high colloidal stability, ultra-small size, low toxicity, high biocompatibility, and size-tunable photoluminescence (PL). 1,2,3,4,5 The ability to tailor their surface offers great advantages to label (bio)molecules of interest and monitor them using advanced imaging techniques. 6 However, Au NCs possess intrinsically quite low quantum yields (QYs); thus, several rational strategies have been developed to tailor and increase their PL properties. 7,8 For example, one strategy involves increasing the organic shell rigidity, which protects the metal core and has led to several-fold PL enhancement in the near-infrared NIR-I (700-900 nm) 9, 10, 11 and NIR-II (900-1700 nm) 12 regions. This PL increase has been associated with the reduction of water molecules in the proximity of the Au NC surface, which minimizes non-radiative recombination processes. 13 A second strategy relies on the ne tuning of the surface chemistry. 14 Millstone's team conducted an extensive study demonstrating the in uence of thiolated ligands 15 on the near infrared emission of ultrasmall metal (gold, silver, copper) nanoparticles. 16,17 In addition, it was shown that the addition of a thiolated ligand is important, and the denticity of the ligand used can in uence the PL band and QY. For example, adding a short dithiol co-ligand to the Au NC surface led to a 6-fold increase in the QY with PL enhancement in the 1100-1300 nm 18 range, which allowed the researchers to monitor them noninvasively in vivo down to a 4 mm depth in mice. 19 Protecting Au NCs within large biomolecules such as proteins can be a rational and elegant strategy for engineering the PL of Au NCs and lead to ideal platforms for biomedical applications. 20 The presence of intrinsic reducing agents and sulfur-containing chemical groups in proteins enables the one-step protection of metal NCs. The synthesis of uorescent protein-directed Au NCs using albumin (BSA, bovine serum albumin) as both a reductant and stabilizer was pioneered by Xie et al. 21 Following this seminal work, various NCs were synthesized with different proteins, showing very high PL brightness and good photostability for biolabeling/imaging purposes.. 22,23,24 Although some studies have shown that proteins retain their structure and biological activity upon protein-directed NC formation, 25,26,27 there is an increasing number of studies demonstrating that the global (and local) structure of proteins is altered after NC formation; these studies were often performed under quite harsh conditions (pH 12, for example), which can also affect the structure of the proteins. 28,29,30,31,32 Additionally, the exact number of metal atoms in protein-directed NCs and their localization within the protein are still under debate, preventing the development of a clear structure-property relationship. 7 It has been recently shown that clusters in proteins may in fact be composed of sub-units of 8 to 10 atoms located in different parts of the protein, and these sub-units can induce partial denaturation. 33,34 Clearly, with such alteration of the native state of the protein host and/or uncontrolled formation and location of clusters within proteins, such nanoclusters may present reduced bioactivity and/or biocompatibility, limiting the widespread use of such materials in biomedical research. Therefore, new synthetic methods preserving the structure and bioactivity of incorporated NCs are in high demand. 35 Ligand exchange plays an important role in thiolated nanoclusters, as the exibility of the gold-sulfur interface enables post-synthetic modi cation. 36 Additionally, functionalization of gold through ligand exchange has been recently exploited for protein carbonylation detection. 37,38 As mentioned before, variations in the nature of ligands and the simplicity of the ligand exchange offer great potential to amplify speci c properties of the clusters, such as the PL properties. 7 In this work, we report for the rst time the incorporation of a single preformed Au 25 nanocluster (Au 25 (pMBA) 18 , pMBA -4-mercaptobenzoic acid) into bovine serum albumin (BSA) protein selectively anchored to one cysteine residue without drastic modi cation of the protein structure. The attachment of a single thiolated Au 25 nanocluster to the protein had a signi cant effect on its NIR-II emission band with a 4.7-fold enhancement of the PL QY. Such water-soluble Au 25 NCs attached to BSA display bright infrared emission between 700 nm and 1300 nm, in contrast to previously reported uorescent BSAdirected Au NCs that only emit below 1000 nm. Mass spectrometry analysis revealed that the attachment of Au 25 NCs to BSA involves ligand exchange between the pMBA ligands and BSA cysteine (Cys) residues. Small-angle X-ray scattering (SAXS) provided an unambiguous localization of a Au 25 NC on the protein via a free cysteine residue, Cys34. To better understand the PL properties of Au 25 nanoclusters attached to BSA, we synthesized mixed liganded nanoclusters, Au 25 pMBA (18-x) Cys x with x=2, 5, 18, through ligand exchange. The origin of NIR-II emission in nanoclusters with mixed ligands determined by optical measurements and density functional theory reveals the key role of the metallic-ligand interface in photophysical pathways. These results pave the way for the tailored synthesis and protein bioconjugation of nanoclusters to extend their uses for diagnostic and therapeutic applications.

Results
Formation of Au 25 pMBA 18 nanoclusters and their attachment to BSA. Au 25 clusters protected by 18 mercaptobenzoic acid (para isomer) ligands were synthesized as described in a previous study. 39 The best results were obtained when a slight excess of Au 25 pMBA 18 over BSA (ratio 1.2:1) was used, but the synthesis was also performed with an excess of protein (ratio 1:3). Upon gentle and continuous stirring for 3 days, the Au 25 nanocluster was incorporated into the BSA protein via ligand exchange between the sulfur groups from pMBA and the sulfur group from cysteine-containing amino acid residues in the BSA.
A schematic representation of the proposed reaction mechanism is shown in Figure 1. These formed Au 25 -BSA conjugates were puri ed using 10% acrylamide SDS-PAGE gel separation. The gel resulted in two bands, which were attributed to Au 25 pMBA 18 and BSA-Au 25 , in contrast with the single band observed for the BSA control (Fig. S1). The slight difference in the positions of the BSA and BSA-Au 25 bands in the gel indicates that both the charge and overall shape of the protein subtly changed due to the incorporation of the preformed Au 25 NCs into the BSA protein. We noticed that Au 25 pMBA 18 could not bind to BSA even when we used an excess of protein (Fig. S1). This result could suggest that Au 25 pMBA 18 is not stuck to the surface but xed inside the protein and cannot react anymore.
Characterization of Au 25 -BSA conjugates. Matrix-assisted laser desorption/ionization time-of-ight mass spectrometry (MALDI-TOF MS) was used to characterize the BSA-Au NCs (Figure 2a). In addition to the signature peak of the BSA protein at 66 kDa, the MALDI spectrum shows an additional peak at higher molecular weight. The molecular weight difference of this peak from BSA was 7.3 kDa, which indicates the attachment of only one gold nanocluster to the protein, as the mass for one Au 25 pMBA 18 nanocluster is 7.68 kDa. However, the observed mass shift of the Au 25 -BSA was lower than expected. This discrepancy can be explained by the removal of 2-3 pMBA surface ligands during incorporation into BSA and the replacement of intrinsically available sulfur groups from the cysteine residues in the BSA. The rst ligand exchange may preferentially occur with only the free cysteine residue of the BSA (Cys34), and then 1 or 2 cysteines in the disul de bridge may become available to bind the Au 25 NCs in the BSA (see Secondary structural changes in the BSA upon Au 25 NC incorporation were measured by circular dichroism (CD) spectroscopy, which is shown in Figure S3. The CD spectra of the BSA and BSA-Au 25 are very similar, suggesting that the BSA-Au 25 retained the α-helicity to a high degree. This observation also con rms the low perturbation of the protein secondary structure content. These results are in stark contrast to the results of Xie's classical method, which was performed under alkaline conditions, and they show the advantage of the synthetic methodology developed here. 21 Small-angle X-ray scattering (SAXS) is a widely used technique for the structural and dynamic characterization of biomolecules in aqueous solutions. 40,41,42 To obtain the scattering pro les of BSA and the BSA-Au 25 complex, inline size-exclusion chromatography coupled to SAXS (SEC-SAXS) was used, as it enables the separation of the major peak corresponding to the BSA monomer from those of the BSA oligomers (Fig. S4a). Consistent with its larger molecular weight, BSA-Au 25 eluted earlier than native BSA. SEC-SAXS frames corresponding to the monomer were averaged using standard procedures for subsequent structural analyses. The SAXS pro les of BSA and BSA-Au 25 present different features along the momentum transfer range measured, indicating that the presence of the Au 25 NC strongly modi es the scattering properties of the protein (Fig. 2b). As expected, BSA-Au 25 presents a slightly larger radius of gyration, R g , than free BSA, 28.16 ± 0.05 Å and 27.95 ± 0.02 and, respectively. When computing the pairwise distance distribution, p(r), we found that the unmodi ed BSA had a D max of 82.1 Å, similar to the value identi ed in earlier studies 43 , and a symmetric p(r) pro le typical of globular proteins. In contrast, BSA-Au 25 had a signi cantly larger D max of 90.85 Å and had a shoulder at small distances, indicating the presence of additional scattering centers in the protein.
Next, the capacity of the crystallographic structure of BSA (PDB entry 3v03) was evaluated to describe both SAXS curves shown in Fig. 2b. While the structure was in excellent agreement with the BSA pro le (χ 2 = 2.2) (Fig. 2b), unsurprisingly, it was unable the describe the curve measured for BSA-Au 25 (χ 2 = 24.6).
To achieve a better t between the experimental data and the theoretical structure of the BSA-Au 25 , we modeled the Au 25 NC (see Computational for Au 25 NC details in the SI) on the surface of the BSA. we selected Cys34 for this model, as it is the only free cysteine in the protein, and we placed its side chain in the position of one of the pMBA molecules of the external protective shell of the Au NC (Fig. 2c). Although the incorporation of the Au 25 NC into the BSA structure improved the agreement with the experimental BSA-Au 25 SAXS curve (χ 2 = 6.55), systematic deviations throughout the momentum transfer range were still observed. A closer inspection of the BSA structure showed a region of the protein (residues 55-119), consisting of a long unstructured loop (residues 105-119), that could be structurally impacted in the presence of the NC. To test this hypothesis, we performed a normal mode SAXS re nement of the BSA-Au 25 structure using the program SREFLEX, 44 allowing movement in this region while maintaining the rest of the protein and keeping the NC xed in place (Fig. 2c).
To show the selectivity of NC binding, we generated seven different structures with the Au 25 NC attached to other cysteines of the BSA (Fig. 2d). Note that these cysteines are engaged in disul de bonds in the native structure. For six of the BSA-Au 25 complexes, the resulting structures displayed worse χ 2 values than the model built using Cys34 as the anchoring point (Fig. 2e). Even after normal mode re nement, all the resulting structures provide worse agreement with the experimental data than the Cys34-anchored structure. In short, the SAXS analyses strongly indicate that the Au 25 NC selectively binds to free Cys34 and that this binding induces structural perturbations to accommodate the metallic cluster.
Optical properties of the Au 25 pMBA 18 and BSA-Au 25 NCs. Figure 3a shows the absorption spectra of Au 25 pMBA 18 and BSA-Au 25 NCs dispersed in water, with both clusters presenting the " ngerprint" band at 680 nm, which is characteristic of [Au 25 (SR) 18 ] − NCs. 45 The absorption spectra of Au 25 pMBA 18 and BSA-Au 25 clusters are similar to the 25-Au-atom cluster (Au 25 (SCH 2 CH 2 Ph) 18 ) with a known crystallographic structure. 46 This result suggests that they all have similar structural features, although UV-vis absorption is not the method of choice to determine or compare structures. The attachment of Au 25 to BSA leads to a signi cant enhancement of the PL signal (upon excitation at 700 nm) at both 850 nm and 1050 nm, as seen in Figure 3b, with a 4.7-fold increase in QY (see Figure 3d). A similar trend is observed for excitation at 400 nm (see Fig. 3c). To con rm that this effect is not only due to simple physisorption of Au 25  The absence of any changes in the PL intensity and lifetime of the Au 25 -BSA complex when added to a concentrated BSA solution is additional evidence for the successful incorporation of the cluster in the BSA protein and thus protection from its environmental surroundings ( Figure S5), indicating the presence of a strong protective shell around the Au 25 in this labeled protein.
In summary, by attaching a single Au 25 NC to BSA through ligand exchange between protecting pMBA ligands and natural sulfur-containing BSA cysteine residues, we managed to tailor the NIR-II signal while keeping the structure of Au 25 intact. Interestingly, the ratio between the relative intensity at 850 nm and 1050 nm strongly depends on the excitation wavelength (compare exc . = 400 nm and exc . = 700 nm, see Fig. 3 and Table S1). The I 1050 /I 850nm ratio is also dependent on the nature of the ligands protecting the Au 25 (for a comparison of Au 25 pMBA 18 and Au 25 -BSA, see Table S1). To better understand the PL properties of the Au 25 nanoclusters that attach to BSA more favorably via cysteine residues, which was suggested by the SAXS/MALDI results, we synthesized mixed liganded nanoclusters, Au 25 pMBA (18-x) Cys x with x= 2, 5, 18, through ligand exchange. Optical measurements and density functional theory allowed us to determine the key role of the metallic-ligand interface in the photophysical pathways.
Insight into the relationship between the structure and optical properties and the origin of the NIR-II emission. The PL of metal nanoclusters originates from a subtle interplay between excitations arising within the metal core and from charge transfer between the metal core and surface ligands. 49,50 A recent study also suggests the in uence of structural distortion accompanied by electron redistribution in photoexcited gold nanoclusters, which induces controllable dual PL emission. 51 However, the detailed PL mechanism of gold nanoclusters is still under debate. 52 The high stability and detailed structure determination of Au 25 (SR) 18 (gold nanoclusters protected by SR thiolate ligands) enabled extensive investigation of the relationship between the structure and optical properties; in particular, it enabled the study of its PL properties. Upon visible excitation, both red (700−800 nm) and near-infrared emissions (approximately and above 1000 nm) have been observed in Au 25 (SR) 18 by different groups. 53,54,55 On the basis of time-resolved emission and nanosecond transient absorption spectroscopy analyses, Meng Zhou and Yongbo Song proposed a simpli ed model, 55 where visible and near-infrared emissions have different lifetimes and arise from the core−interface charge transfer state and the Au 13 core state, respectively.
When Au 25 is incorporated into BSA, at least one ligand exchange with the protein occurs via a cysteine residue, which signi cantly modi es the PL pro le (Fig. 3b). To reveal the key role of the metallic-ligand interface in the photophysical pathways of BSA-Au NCs, we produced Au 25 pMBA (18-x) Cys x NCs with x=0, 2, 5, and 18 by ligand exchange (between pMBA and cysteine) from the Au 25 pMBA 18 precursor nanocluster. These NCs were fully characterized by ESI-MS (see Fig. S6). Au 25 pMBA (18-x) Cys x with x=0, <2>, <5> and 18 presents the typical absorption features of Au 25 NCs with the characteristic band centered at 680 nm and a tail band above 780 nm (Figure 4a). 56 The PL pro le of these Au 25 NCs dispersed in water exhibits two main broad bands at 800-950 nm and 1050-1250 nm (Fig. 4b and c). When the number of Cys molecules on the Au NC surface increased from 2 to 5, an increase in the PL band at 1050 nm was observed (see Fig. 4b and c). To compare the evolution of the two main NIR-II emission bands, we determined the PL ratio of the peak at 1050 nm to that at 920 nm to determine if there is a correlation between the presence of the co-ligand and the PL enhancement at 1050 nm (Table  S1). The results indicated a more pronounced effect when 700 nm was used as the excitation wavelength than when 400 nm was used, indicating that the 1050 nm emission probably involves photophysical relaxation pathways taking place within the metal core. 55 Indeed, with an excitation wavelength of 700 nm, the lowest energy excited states (S1, S2) are mainly involved in photoexcitation, as con rmed by the time-dependent density functional theory (TD-DFT) linear absorption spectrum ( Fig. 4d and 4f, and see computational details in SI). Low-lying states in the NIR are key to obtaining large two-photon absorption cross sections (due to double resonance effects), which would make them ideal labels for multiphoton excited luminescence (see Fig. S7). The lowest excited states in the absorption spectrum mainly belong to the "core" in Nature, primarily arising from transitions from the occupied P orbitals into the rst and second sets of D orbitals (see Fig. 4e and Table S2). For low-lying excited states, some contributions of the ligands to the transitions are more readily observed in Au 25 pMBA 16 Cys 2 than in Au 25 pMBA 18 (Table   S2). Under 400 nm excitation, the nature of excited states is both "interface-like" and "core-like." The characteristic of such excited states is mainly "interface-like" 57 for which the contribution from the Au-S interface in orbitals is more pronounced, particularly in the cysteine-containing Au NCs (Table S2).
Although pMBA and cysteine are rich electron donors, cysteine is more exible due to the Cα-Cβ bond and is better able to interact with the surface of the gold core than pMBA, as its carboxylic group points outward toward the surface of the gold core. Such electron-rich donor groups may contribute to the "interface-like" excited states and thus may increase the contribution of the 850 nm band. 52 This phenomenon is particularly true for Au 25 Cys 18 , which mainly displays a strong emission at 850 nm.
On the basis of these experimental and theoretical results, we propose the following mechanism to explain the relative contribution of emission bands (at 850 and 1050 nm) of the Au 25 -BSA conjugates ( Figure S8): rst, the local environment of Au 25 NCs attached to BSA is more rigid than free Au 25 pMBA 18 and prone to multiple energy transfers associated with intersystem crossing (reinforced ISC), which could explain the overall enhancement in PL emission and the longer PL lifetimes. Second, the nature of ligands, and in particular their capability to interact through electron-rich donor groups with the surface of the Au core, may increase the number of surface states involved in excitation (speci cally at 400 nm).

Discussion
Here, we present a simple route to attach individual Au 25 NCs to a BSA protein via a cysteine residue.
SAXS and molecular modeling provide a clear localization of Au 25 within the protein (Cys 34), and MALDI-MS con rms that ligand exchange with a cysteine residue of BSA occurs on the gold nanocluster surface. The change in the local microenvironment of the Au NCs incorporated into the BSA boosts the NIR-II signal but preserves the structure of the Au 25 . Rigidi cation of the surface molecules within the BSA protein reinforces the intersystem crossing pathway involved in photoluminescence. Cysteine ligands exchanged during the attachment of Au 25 to BSA allow for an increased contribution of core−interface charge transfer states. Both effects induce enhancement and shift the photoluminescence signal. In addition to the importance of keeping the structure of Au NCs unchanged within the protein, which enables us to decipher the structure-property relationships, it is of interest to use 700 nm excitation for in vivo optical imaging and photoactivation, both in the linear and nonlinear optical regimes. This study paves the way to designing new photoemitters with tunable NIR-I/NIR-II emission via visible or NIR excitation. Finally, engineering mutations in proteins with cysteines will allow for more selective control of the position of atomically precise NCs in proteins and of the number of ligands exchanged, opening new routes for extending their uses in diagnostic and therapeutic applications.  Figure 1 Schematic illustration of the reaction process leading from the formation of Au25 nanoclusters to their attachment to BSA.

Figures
Step I. Formation of thiolated Au25 NCs.
Step II. Ligand exchange with cysteine residues in the BSA leading to the anchoring of Au25 nanoclusters to BSA. Cysteines are highlighted in red, and the Au25 NC structure is shown anchored to Cys34. The SREFLEX re ned model of the Au25-BSA (light blue) superimposed on the crystal structure of BSA. The position of the part that was allowed to be moved by SREFLEX is shown in darker shades of the same colors in both structures. d) Two views of the BSA conjugated to Au25 at various cysteines, tested to determine the agreement with the SAXS data. e) The χ2 values between the theoretical SAXS pro le of BSA conjugated to Au25 at different cysteines calculated directly from the structure using Crysol (gray bars) or after re nement using SREFLEX (red bars).

Figure 3
a) Absorption spectra of Au25pMBA18 and BSA-Au25 in water. PL spectra of Au25pMBA18 and BSA-Au25 in water, with (b) exc. 700 nm b) and c) exc. d) Table 1: Relative QY of Au25pMBA(18-x)Cys-x and Au25-BSA at 700 nm excitation using IR125 as a reference. e) PL lifetime decay of Au25pMBA18 in water, Au25pMBA18 with BSA, and BSA-Au25 in water ( exc. 485 nm; em. 920 nm).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.