Introduction

The field of photoluminescent noble metal clusters is a hot topic in materials science [1]. Subnanometer-sized gold particles show photoluminescence emission that can be tuned in a wide range of the spectrum, from ultraviolet through visible to the near infrared [2]. These clusters show good photostability compared to organic fluorophores [3]. Furthermore, unlike semiconductor quantum dots [4], gold clusters do not have the disadvantage of susceptibility to oxidation.

Gold quantum clusters (AuQCs) can be prepared in a variety of ways. A main strategy for preparation is a top-down approach in which gold clusters are prepared by controlled etching of monolayer-protected gold nanoparticles [5–8]. A ligand exchange-based reaction has also been reported [9]. Bottom-up techniques are also known; polyvinylpyrrolidone [10] and polyamidoamine dendrimers are suitable protective agents for cluster production [2]. In the latter two methods, sodium borohydride is used for the reduction of AuIII. Photoluminescent metal clusters with small ligands are already applied for cellular imaging [5, 11, 12].

Alternative bottom-up methods are being developed for cluster preparation: recently, some water-soluble proteins (will be referred to as active clustering proteins, ACPs) have been reported to react with HAuCl4 to produce red-emitting AuQCs in situ in aqueous phase, acting as reducing and complexing/protecting agents [13]. These are bovine serum albumin (BSA), trypsin, pepsin, horseradish peroxidase, lactoferrin, human serum transferrin, lysozyme, and cellular retinoic acid-binding protein 2 (CRABP2) [14–21]. Protein-conjugated gold clusters are already applied as analytical sensors of Hg2+, Pb2+, Cu2+ ions, and glutathione [15, 21–24]. Precursors used in protein-based techniques are nontoxic; moreover, BSA-protected clusters do not show significant toxicity in vitro [25], which allows their application in biological systems. In the future, expansion of the protein-directed production methods can be expected.

Although there is a growing number of detailed works in the field of protein-mediated AuQC preparation, it is yet unclear which physicochemical characteristics of proteins are crucial in cluster formation, and there are no general criteria for protein selection in the literature. Furthermore, the design of peptides for the synthesis of gold nanoparticles has been elaborated [26], but there is no such work in the literature for AuQC production.

In an earlier study, cysteine and tyrosine residues were claimed to be responsible for the cluster forming activity of proteins [14], but there is a lack of experimental evidence for this theory. Nevertheless, investigation on the oxidation state of gold suggested that AuI and Au0 could be both present in the clusters [14–16, 18], which implicates that different residues are involved in the process. It has been demonstrated that cysteine is not the only amino acid by which AuI species can be formed: methionine can also reduce gold salts in a similar manner [27, 28]. Tryptophan, having a similar redox character to that of tyrosine at high pH values [29], has been found to be the strongest reducing amino acid for AuIII in its free form and also as a peptide residue for gold nanoparticle synthesis [26]. Tyrosine, on the other hand, could be essential, since it could reduce AuIII to unstable AuII species [30], which can be an important step in the reaction. Moreover, cysteine has been reported to facilitate the formation of AuI thiolate polymers [31], which are responsible for the attractive forces in clusters through aurophilic interaction [32–35].

Peptides can also be used for cluster preparation. Insulin, for example, has been shown to produce AuQCs for cellular imaging [36]. The simplest templating peptide is glutathione (γ-Glu-Cys-Gly), which can effectively stabilize subnanometer clusters if an external reducing agent is also used (note that glutathione itself cannot reduce AuIII to produce Au0) [7, 33]. Glutathione-protected clusters, however, have low photoluminescence quantum yield compared to that of protein-coated clusters [37], which is a disadvantage in imaging techniques. In the literature, AuQC synthesizing activity of peptides has not been widely investigated. To date, only a single artificial peptide (CCYRGRKKRRQRRR) has been designed especially for AuQC production [3].

Another key aspect is the concentration of the protein necessary for the reaction. It has been confirmed that red-emitting Au25 clusters consist of a 13-atom core, surrounded by 12 peripheral gold atoms forming an AuI-S semi ring [38–40]. The existence of the latter requires a minimum number of sulfur-containing residues per gold cluster [39, 41]. These considerations suggest the existence of a critical initial molar ratio of the protein and the gold precursor necessary for cluster formation. Although it has been observed that cluster formation takes place if the concentration of the protein is relatively high [14, 18, 21], such critical initial molar ratio has not yet been defined in the literature.

In our study, we intended to identify the criteria that determine cluster-producing feature of proteins and peptides. We considered the following questions: (1) Is the presence of cysteine and tyrosine in a protein or peptide sufficient? (2) If cysteine and tyrosine do have a key role in cluster production, can either of them be replaced with residues of similar character? (3) Do ACPs show similar motifs in primary structure? (4) Can AuQCs be prepared using sulfur-containing and reducing amino acids in their free form, without the use of an external reducing agent? Experiments have been performed using amino acid mixtures, peptides, and proteins, respectively. Furthermore, sequences of ACPs have been analyzed.

Three systems were studied: proteins, peptides, and amino acid mixtures, respectively. In the case of protein-mediated AuQC production experiments, we have chosen commercially available proteins in a molecular weight range of 30 to70 kDa. BSA, porcine trypsin, bovine hyaluronidase-3, and concanavalin A were selected in order to clarify if protein properties (number of cysteine residues per molecule, percentage of disulfide-bonded cysteine residues) affect their efficiency in cluster formation. Furthermore, hyaluronidase-3 is a water-soluble protein with sulfur-containing and reducing residues, and its cluster-producing feature has not been investigated earlier. Since known ACPs show diversity in the secondary and tertiary structure, we proposed that cluster-producing activity of proteins is determined mainly by their amino acid profile and/or primary structure. Sequence of all proteins used in our experiments and ACPs in the literature have been analyzed to study the function of the primary structure in cluster formation.

Peptides have been designed for our experiments to elucidate if either cysteine or tyrosine can be substituted in the reaction (cysteine with methionine and tyrosine with tryptophan) and to clarify if the position of cysteine residues affects cluster formation. Gonadotropin-releasing hormone derivatives as potential drug-targeting moieties suitable for targeted cancer therapy and cancer detection were also chosen to these studies [42]. These compounds contain cysteines with free thiol group or in disulfide bridge, as well as both tyrosine and tryptophan.

For our experiments with amino acid mixtures, we have chosen sulfur-containing and strong reducing amino acids, proposing that their presence is sufficient for cluster formation. It has been reported that the former can protect clusters and form AuI complexes [28, 37–41], while the latter can produce Au0 [14, 26, 29]. We designed our mixtures based on the assumption that tyrosine and cysteine play the key role in the process [14], but also investigated the possibility of synergic effects with tryptophan and methionine. Mixtures of W-M and W-C have not been studied, since they have been found inefficient in AuIII reduction earlier in the literature [26].

In this paper, we specified certain crucial parameters whereby an easy selection or design of an appropriate host biomolecule and the adjustment of the precursor concentration for the green production of photoluminescent gold clusters will be possible.

Experimental

Materials

Gold(III) chloride hydrate, 99.999 %, was purchased from Aldrich. BSA (≥98 %), bovine hyaluronidase-3 (H3), concanavalin A (cA), porcine trypsin, l-tyrosine (≥99.0 %), l-methionine (99.0–101.0 %), l-cystine (≥99.5 %), and d,l-dithiotreitol (≥99.0 %) were ordered from Sigma. l-cysteine (99 %) and l-tryptophan (99 %) were obtained from Alfa Aesar. Phosphate-buffered saline tablets (for buffer solutions of pH 7.4) were purchased from Fluka. Sodium hydroxide solution (1 M), analytical grade, was purchased from Carlo Erba. Model peptides CCYE6R3, CCWE6R3, and MMYE6R3 (99 %) were ordered from GenScript. Model GnRH peptide derivatives, with >95 %, were prepared at the Research Group of Peptide Chemistry at Eötvös Loránd University, Budapest (see Online Resource 1). Deionized water of Millipore purity (18.2 MΩ cm) was used for all experiments.

Production of AuQCs with proteins

Four different proteins were used in the experiments, properties of which are summarized in Table 1. Protein-coated gold clusters were prepared based on previously reported methods [14, 15]. Briefly, aqueous solutions of the proteins (25 mg mL−1 for BSA, 40 mg mL−1 for trypsin, 50 mg mL−1 for hyaluronidase-3, and 100 mg mL−1 for cA, respectively) were made. Aliquots were then mixed with equal volume of HAuCl4 solution (10−2 M). After 10 min of stirring, 1 M NaOH solution (5 % of the original mixture in volume) was injected to set the pH to 12.0. Mixtures were reacted at 40 °C overnight. Products were stored in the refrigerator at 5 °C.

Table 1 Properties of proteins used in AuQC production experiments
Full size table

In the case of proteins, we defined a critical molar ratio, ϕ c, which is the lowest protein/gold ratio in the reaction by which photoluminescent gold clusters form. In order to determine ϕ c for H3, the concentration of HAuCl4 was kept constant (5·10−3 M), experiments were carried out at protein/gold molar ratios of 0.180, 0.090, and 0.045, respectively. We proposed that reactions with ratios below ϕ c result in the formation of nanoparticles similarly as reported earlier for other proteins [14, 21]. We estimated ϕ c considering the lowest protein concentration yielding photoluminescent clusters and compared our results with data extracted from the literature.

Preparation of AuQCs with peptides

Amino acid sequences of the peptides used in our experiments are shown in Table 2. The concentration of peptides in the reaction was constant (3.6·10−4 M), peptide/gold ratio was adjusted to 1.00. In the case of CCYE6R3, a peptide gold ratio of 0.36 was also tested. All experiments were carried out at pH 12.0, at 40 °C, stirring the mixtures overnight.

Table 2 Amino acid sequence of peptides and peptide dimers used in Au quantum cluster synthesis experiments
Full size table

AuQC production experiments with amino acid mixtures

HAuCl4 was reacted with mixtures of different amino acids to clarify if free amino acids can produce gold quantum clusters. Concentrations and ratios of different amino acids were adjusted to mimic the BSA-mediated synthesis (Table 3). In these experiments, NaOH was added prior to the addition of the HAuCl4 solution (10−2 M, equal volume to that of the amino acid solution) to dissolve nonpolar amino acids (tyrosine, methionine, and tryptophan) efficiently at pH 12.0. Analogous experiments have also been carried out using l-cystine in the mixtures.

Table 3 Composition of amino acid mixtures used in gold cluster production experiments
Full size table

Characterization techniques

Samples containing proteins and free amino acids were diluted with phosphate-buffered saline (pH = 7.4, SI = 162.7 mM) to ten times of their original volume. The pH of the mixtures was measured with a Jenway 3540 pH meter. Mixtures containing peptides were studied undiluted in order to achieve sufficient signal-to-noise ratio in photoluminescence (PL) measurements. Steady-state PL spectra have been collected using a Shimadzu RFPC 3501 fluorimeter. Samples were excited at 350 nm, excitation and emission slits of 5 nm were used. We detected the presence of AuQCs by measuring spectral properties of the reaction products.

It is well known that the luminescence lifetime red-emitting gold clusters is in the range of microseconds [37, 38, 43]; therefore, our samples showing red emission have been characterized by measuring luminescence lifetimes to confirm the presence of gold clusters. Time-resolved photoluminescence (TRPL) measurements have been carried out using an Edinburgh Instruments FLSP920 luminescence lifetime spectrometer. The excitation light source was an EPL-470 diode laser (473 nm; pulse duration, 140 ps at FWHM) and the detector was a Hamamatsu R3809U-50 microchannel plate photomultiplier. Decay curves were analyzed by nonlinear least-squares fitting method using Edinburgh F900 software. UV–visible absorbance spectra were obtained with a ThermoSpectronic Unicam UV500 spectrophotometer. Structural changes of proteins were followed by Fourier transform infrared (FTIR) spectroscopy, by observing the shifts of amide I and amide II bands. FTIR-attentuated total reflectance (ATR) spectra have been collected using a Varian FTS-2000 FTIR Spectrophotometer with a Golden Gate ATR head.

Comparison of ACP structures

Protein sequences were obtained from the UniProt database (https://www.uniprot.org), amino acid profile of all proteins can be found in Table 6 (see Appendix). Sequences were aligned by the Needleman–Wunsch algorithm using an online alignment tool (http://www.ebi.ac.uk/Tools/psa/emboss_needle). The BLOSUM60 matrix was applied for the calculation of scores. The score of the alignment of human transferrin and bovine lactoferrin was used as a reference value, since these proteins are structural analogues (https://www.uniprot.org).

Results and discussion

Production of AuQCs with proteins

In order to determine which criteria are crucial in the case of proteins to make them efficient in cluster production, we investigated sequences of known ACPs. We have found that primary structures of ACPs do not have any ubiquitous sequential motif (see Online Resource 2). This suggested that clustering efficiency of a protein is not determined by specific motifs in its sequence, but the presence of certain residues is crucial. Considering the amino acid profiles of the investigated proteins (Table 3, Appendix), it is also apparent that there are ACPs that have no histidine (CRABP2), methionine (insulin), or tryptophan (egg white lysozyme) residues, which suggests that the presence of these amino acids is not essential in cluster formation. However, tyrosine and cysteine residues are present in the primary structure of all ACPs.

Steady-state PL emission spectra of reaction mixtures of HAuCl4 and H3, and cA, are shown in Fig. 1. According to PL measurements, three of the selected proteins (BSA, trypsin, and hyaluronidase-3) efficiently produced AuQCs, while concanavalin A, the one with no cysteine residues, was found to be inactive in cluster synthesis. BSA- and trypsin-coated clusters showed similar emission properties to those earlier reported in the literature [14, 15].

Fig. 1
figure 1

Emission spectra of reaction products of HAuCl4 with protein: BSA (solid line), bovine hyaluronidase-3 (dashed line), porcine trypsin (dotted line), and concanavalin A (dash–dot line), respectively (an excitation wavelength of 350 nm was used)

Full size image

H3-protected AuQCs showed an emission peak at 615 nm. Red emission was not detected in the case cA, and no nanoparticle formation was observed either. Emission lifetimes of samples with red emission have been measured; two lifetime components (0.9 and 2.3–2.4 μs) were observed. Results of TRPL measurements in case of products prepared using proteins are summarized in Table 4.

Table 4 Emission lifetime values (τ) and pre-exponential constants (α) of red-emitting AuQCs produced using proteins
Full size table

As lifetimes of microsecond range are characteristic for thiolic ligand-protected gold clusters [37, 38, 43], these lifetime values indicate cluster formation unambiguously. TRPL results thus provided evidence that H3 is also an ACP. Since BSA, trypsin, and H3 have different SH/disulfide ratios, as shown in Table 1, we can conclude that cysteine can be effective in thiol and also in disulfide form. In the case of porcine trypsin, however, we measured low emission intensity, probably due to autolysis of the enzyme.

In the case of H3, different protein/gold ratios were used in AuQC preparation experiments. We hypothesized that the value of ϕ c is solely determined by the number of cysteine residues in a protein, considering a stoichiometry of 18 thiol groups and 25 gold atoms [39, 40]. Based on this assumption, the value of the critical ratio must be in inverse relation to the number of cysteine residues in a single molecule.

$$ {\phi}_{\mathrm{c}}\approx \frac{0.72}{n_{\mathrm{Cys}}} $$
(1)

where n Cys is the number of cysteine residues in a single molecule. We supposed that ϕ c can be estimated by lowering the concentration of protein in the reaction mixture while keeping the concentration of HAuCl4 constant.

We observed that the concentration of H3 is a key factor in AuQC synthesis. UV–visible absorbance and PL spectra of the synthesis mixtures of H3 and HAuCl4 are shown in Fig. 2.

Fig. 2
figure 2

a Emission spectra of reaction products of H3 and HAuCl4 with protein/gold ratios of 0.180 (solid line), 0.090 (dashed line), and 0.045 (dotted line), respectively (an excitation wavelength of 350 nm was used); inset shows UV–vis absorbance spectra of the samples; b photograph of samples taken under UV lamp (365 nm)

Full size image

Results show that the nature of the reaction products does depend on the protein/gold ratio in the case of H3; nanoparticles or clusters can form depending on the synthetic conditions. At lower protein/gold ratios, non-fluorescent nanoparticles formed, plasmon absorption peaks of which were observed in the UV–Visible absorbance spectra. To compare our experimental results with data extracted from the literature [14, 18, 21], we plotted the estimated ϕ c values for different proteins, and the theoretical values calculated based on the cysteine content of the proteins (using Eq. 1) are demonstrated in Fig. 3. Theoretically predicted values of ϕ c for proteins using the Au25SR18 stoichiometry in the calculation [39, 40] and experimentally estimated values proved to be coherent, indicating that ϕ c is in inverse ratio to the number of cysteine residues in a protein. Moreover, we found that disulfide-bonded and free cysteine residues can be considered as equivalent.

Fig. 3
figure 3

Theoretical ϕ c values calculated from the number of cysteine residues in a single molecule (Eq. 1) (blank circles) and estimated ϕ c values for different proteins (solid circles): bovine hyaluronidase-3 (solid circle, based on our measurements), BSA (star [14]), bovine lactoferrin (square [18]), and egg white lysozyme (triangle [21]), respectively

Full size image

Preparation of AuQCs with peptides

In the case of model peptides (Table 2), CCYE6R3, [d-Cys6]-GnRH-II, and ([d-Cys6]-GnRH-II)2 produced AuQCs in case of peptide/gold molar ratio of 1:1. PL spectra of reaction mixtures are shown in Fig. 4.

Fig. 4
figure 4

Emission spectra of reaction products of HAuCl4 with CCYE6R3 (solid line), CCWE6R3 (dash–dot line), MMYR6R3 (dash–dot–dot line), [d-Lys6(Cys)]-GnRH-I (short dash–dot line), [d-Cys6]-GnRH-II (dashed line), and ([d-Cys6]-GnRH-II)2 (dotted line), respectively (an excitation wavelength of 350 nm was used)

Full size image

As in the case of proteins, emission lifetimes of red-emitting samples have been measured, except for [d-Cys6]-GnRH-II disulfide, since the reaction product formed an insoluble precipitate. Emission lifetimes of our samples with red emission are demonstrated in Table 5.

Table 5 Emission lifetime values (τ) and pre-exponential constants (α) of red-emitting AuQCs produced using peptides
Full size table

Microsecond lifetimes have been observed in the case of the reaction mixtures using model peptides as reactants, confirming the formation of AuQCs. Replacement of cysteine with methionine led to the inefficiency of the peptide. We observed the same effect when tyrosine was replaced by tryptophan. These results revealed that cysteine and tyrosine are exclusively responsible for cluster formation and cannot be substituted by other residues, which is an experimental evidence for earlier assumptions [3, 14]. Cysteine might be essential in cluster formation due to the electron withdrawing effect of the thiol–gold bond, which plays a crucial role in cluster stabilization [44]. If CCWE6R3 was used as a reactant, addition of 1 molar equivalent of free tyrosine or MMYE6R3 to the mixture did not facilitate the formation of AuQCs, emphasizing the importance of intramolecular processes in the cluster formation reaction. Cluster synthesizing potency of [d-Cys6]-GnRH-II and ([d-Cys6]-GnRH-II)2 showed that non-terminal cysteine residues are active . On the other hand, the inefficiency of [d-Lys6(Cys)]-GnRH-I, [d-Lys6(Ac-Cys)]-GnRH-I, and their disulfide dimers indicated that the position of Cys residues is a crucial factor in AuQC production. There was no difference in cluster formation using acetylated or non-acetylated versions of GnRH-I derivatives. For this reason, we can conclude that the free amino group on the terminal cysteine is not the key factor in prevention of AuQC production. We presume that our conclusions for peptides are also valid for proteins.

FTIR spectra of the reaction products prepared using proteins and peptides

FTIR-ATR spectra and the corresponding second derivative spectra, demonstrated in Fig. 5, show shifts of the amide I and amide II bands to higher wavenumbers in the case of red-emitting samples compared to native proteins.

Fig. 5
figure 5

FTIR-ATR spectra (solid lines) and second derivatives of the corresponding spectra (dotted lines) of native biomolecules and products of HAuCl4 reacted with a proteins and b peptides

Full size image

In the spectra of the negative controls, cA and [d-Lys6(Ac-Cys)]-GnRH-I, shifts of the corresponding bands were negligible. Shifts of the amide bands are due to conformational changes of the biomolecules, as it has been demonstrated earlier [44]. These results indicated that the proteins go through conformational changes during the reaction, which are associated with the presence of gold particles, and confirmed the formation of subnanometer entities. Changes in the secondary and tertiary structures of several ACPs have been detected earlier [14, 15, 17, 19, 34, 45, 46], which is consistent with our results.

AuQC production experiments with amino acid mixtures

In the case of amino acid mixtures, the reaction with HAuCl4 did not result in the formation of either red-emitting clusters or nanoparticles. However, a strong fluorescence band was observed at 400 nm, which indicated the formation of dityrosine [47–49] as a result of a redox reaction between AuIII and tyrosine (dityrosine was not detected if HAuCl4 was not added to the amino acid mixtures). One can find this intriguing since cysteine and methionine can form highly stable complexes with gold, limiting the reducibility of AuIII in the case of tryptophan [26]. Considering these, our results suggested that tyrosine can react with gold species even if they are in complex with sulfur-containing residues, which can be an explanation for our observation that replacement of tyrosine with tryptophan can hinder the clustering efficiency of peptides. A possible reason for the inefficiency of free amino acids under our experimental conditions is that they cannot provide nucleation sites for gold species: in a proposed mechanism, formation of gold nuclei was assumed to be a crucial reaction step [34]. Cysteine-protected Au25 clusters can be prepared using an organic phase method, in which aggregation sites are established using micelles of cethyltrimethylammonium bromide [50].

Conclusions

In order to rationalize the criteria for the selection and design of bioreagents for AuQC synthesis, we performed experiments with amino acid mixtures, model peptides, and proteins, and investigated the primary structures of proteins. Studies on peptides showed that the coexistence of cysteine and tyrosine in peptides can determine if a biomolecule can potentially mediate the formation of photoluminescent, red-emitting gold clusters. These observations provide experimental evidence for earlier assumptions [3, 14]. Furthermore, our results showed that effectiveness of cysteine residues is not hindered by their disulfide-bonded state. On the other hand, their location in the peptide sequence is crucial. However, further studies with model peptides are required in order to clarify the effect of the intramolecular distance between cysteine and tyrosine residues.

We also observed that AuQCs cannot be synthesized with mixtures of free cysteine and tyrosine, which indicated that proteins and peptides do not only play a role in the complexation and reduction of gold species but also provide anchoring sites for gold nuclei [34].

Our experiments also led to the conclusion that a single cysteine residue can make a protein or peptide an efficient clustering molecule, but the concentration of the protein determines if AuQCs or Au nanoparticles are produced. We confirmed the existence of a critical protein/gold ratio ϕ c, above which red-emitting clusters form. The value of ϕ c, however, is in inverse relation to the number of cysteine residues in a molecule.

As a consequence of our investigation, a high variety of conjugates of photoluminescent gold clusters and proteins can be synthesized via a green synthesis route and can be designed for potential fields of application.