Reaction of Native HuHf with AuTM and Biophysical Characterization of the HuHf@AuTM Adduct

Intact HuHf was dissolved in PBS and then reacted by simple incubation with increasing amounts of AuTM (to achieve AuTM:nanocage molar ratios of 24 : 1, 72 : 1, 120 : 1, 240 : 1, 480 : 1 and 720 : 1). The interaction between apo-HuHf and AuTM in solution was investigated by ESI MS measurements. The spectra of each sample were recorded after 3 hours of incubation followed by extensive dialysis. The ESI MS spectrum of native HuHf shows an intense peak with a molecular mass of 21093 Da, which can be readily assigned to the monomeric species of the protein; indeed, it is well known that the ferritin nanocage is broken down into individual subunits during the ESI process.²² The complete titration profile is shown in Figure 2. Interestingly, the addition of increasing amounts of AuTM induces the progressive appearance of a new peak at 22057 Da corresponding to a protein adduct carrying a metallic fragment of mass 935 Da per subunit. It is noteworthy that, as the titration proceeds, the peak of the adduct continuously increases in intensity, while the peak of the native apo-subunit decreases until it disappears.

We interpret this spectral behavior as the progressive and complete conversion of the native protein into a single type of HuHf@AuTM adduct. The binding process appears to be highly cooperative: indeed, no other adducts showing different mass increases were detected during the titration. We then attempted to understand the nature of the protein bound fragment of mass 935 Da. This increase in mass is consistent with the binding of a metallic fragment with the general formula Au₃TM₂Na₂ (theoretical mass of 935.1619 Da). The characteristic profile of the ESI MS titration inspired the experimental procedure to prepare the HuHf@AuTM bioconjugate. To this end, the HuHf nanocage was dissolved in PBS buffer and then reacted with a large molar excess of AuTM (at an AuTM:HuHf nanocage ratio of 720 : 1, which corresponds to an AuTM:subunit ratio of 30 : 1), for 3 hours at room temperature (RT), pH 7.4, followed by repeated dialysis against buffer. The gold content per cage was quantified by ICP measurements. An aliquot of the sample was subjected to buffer exchange to make the sample suitable for ESI MS analysis. The resulting ESI MS spectrum is fully consistent with the final spectrum of the above titration, demonstrating the complete conversion of apoferritin into the bioconjugate. No signal for apo-HuHf is detected, indicating that the bioconjugate is stable to extensive dialysis and no gold release occurs. ICP results indicate an average of ~70 gold atoms per HuHf cage (i.e., ~2.9 gold atoms per subunit). Notably, this experimental procedure was highly reproducible and constituted the standard protocol for the preparation of the HuHf@AuTM bioconjugate throughout this study.

ESI MS studies previously performed on a number of model proteins suggest that gold compounds preferentially bind proteins at the level of free cysteines.^23-26 The exclusive binding of AF to the free cysteines of human ferritin (i.e, Cys90, Cys 102 and Cys130, Figure 1D) was clearly documented in our previous study.¹⁰ We thus hypothesized that the binding of AuTM to apoferritin might also occur at the level of the free cysteines of the protein. To validate this hypothesis, some mutants of HuHf were used in which the three free cysteines of the protein (Figure 1D–F) are replaced by Ala residues. Specifically, the following HuHf mutants were used for AuTM binding studies: the single mutants C90A and C130A, and the double mutant C90AC102A. As in the case of wild type apoferritin, these mutant ferritins were challenged with a saturating amount of AuTM for 3 h at RT and the resulting ESI MS spectra were recorded after extensive dialysis (Figure 3). The corresponding titrations are shown in Figure S1 and Figure S2. The ESI MS spectra highlight remarkable differences in the respective reaction profiles. In fact, the C90A and C90AC102A mutants show an almost complete or even complete loss of the ability to bind gold, clearly confirming the pivotal role of cysteines 90 and 102 as exclusive anchoring sites for gold atoms. In the case of C90A a very small peak is visible at +1083 Da (Figure 3A and B).

Conversely, the reaction with the C130A mutant led to complete conversion of the native protein into a metalated species (again with ICP indicating an average of ~2.9 gold atoms per HuHf subunit). Notably, for the latter variant, we detected the presence of a single ESI MS peak for the resulting adduct with a mass of 22146 Da, corresponding to full protein metalation, involving a fragment with a mass of 1083 Da per monomer (Figure 3C). Through spectral simulation, the metallic fragment was found to correspond to the Au₃TM₃Na₂ moiety with a theoretical mass of 1083.8518 Da. This fragment thus appears to contain an additional thiomalate ligand compared to the 935 Da fragment detected in the interaction of AuTM with the wild type apoferritin. This might be due to a partial destabilization of the subunit structure of the C130A variant under the ESI conditions, causing Cys90 to move away from Cys102, and then favouring the binding of an additional TM moiety in the place of a protein Cys to one of the two terminal gold(I) ions.

In order to further characterize the HuHf@AuTM system at the molecular level, the samples of two bioconjugates, i.e., the adduct of AuTM with the wild type protein (HuHf@AuTM) and the corresponding adduct with C130A HuHf, were trypsinized according to a standard procedure²⁷ and the tryptic fragments analyzed by ESI MS in comparison with apo wild type HuHf.

All the expected tryptic fragments (T1-T20) were detected for apo- HuHf and its bioconjugates, with the exception of fragments T3, T4 and T18 (Figure 4). Both Cys90 and Cys102 belong to fragment T12. On trypsinization of either WT or C130A HuHf@AuTM, we detected a peak corresponding to fragment T12 with a mass increase of 1083 Da. This corresponds to the peak observed for the intact C130A adduct (Figure 3C); presumably, the loss of the three-dimensional structure that likely occurs in the tryptic fragment vs the intact subunit causes an increase in the distance between Cys90 and Cys102 permitting the binding of the Au₃TM₃Na₂ moiety. Notably, no metalation was observed on the tryptic fragment T15, which contains residue 130. These observations suggest that metalation occurs exclusively at the level of the portion of the subunits containing the cysteine residues 90 and 102, i.e., the tryptic fragment T12.

The combined information arising from the ESI MS and ICP measurements performed on the wild type and mutant HuHf@AuTM conjugates as well as from the trypsinization experiments allowed us to develop a computational model of the 3D structure of HuHf@AuTM (Figure 5). Gaussian calculations²⁸ for Au₃TM₂(CH₃S)₂ and Au₃TM₃(CH₃S) yielded structural parameters (bond lengths, S−Au−S angles) that closely matched those observed in the X-ray structure,³ although we did not compute a polymeric structure, but only a fragment. Then, the CH₃S moieties of Au₃TM₂(CH₃S)₂ were replaced by Cys90 and Cys102 to model the species observed for WT HuHf@AuTM.

Similarly, the CH₃S moiety of Au₃TM₃(CH₃S) was removed to establish a covalent bond with Cys102 to model the minor species observed in the C90A HuHf@AuTM adduct. MD simulations were analyzed in terms of the overall structural stability of the adducts, measured by the displacement of the backbone coordinates of each subunit as a function of the simulation time, as well as the fluctuations of the inter-subunit contacts, which define the quaternary structure of the nanocage.

The imposition of coordination bonds does not alter the tertiary and quaternary structure of the cage. Interestingly, the dynamics of the C90A HuHf adduct as measured by the mean displacement of the Au atoms over time with respect to their average position is comparable to that of HuHf@AuTM despite the latter having one more coordination bond with the protein.

In our structural model (Figure 5) of the adduct a fragment of the type Au₃TM₂ is associated with the protein through the formation of coordinative Au−S bonds to Cys90 and Cys102. When one of the two Cys is removed (e.g., in the C90A variant) or when the subunit fold is destabilized, thus affecting the distance between C90 and C102 (e.g., in the T12 fragment resulting from trypsinization), we observed an Au₃TM₃ fragment with a free end and a single Cys-anchor, instead of the Au₃TM₂ fragment with two-Cys anchoring points on the subunit surface. The lack of metalation on Cys130 is consistent with its buried location inside the 3-fold channels (Figure 1E). These channels facilitate the selective diffusion of small cationic species,^{29, 30} but are not suitable to host large anionic metal fragments like AuTM.

Antiproliferative and Metabolic Effects of the Adduct vs the Free Drug

The actions of AuTM and of HuHf@AuTM were comparatively investigated in a cellular model of human ovarian cancer, i.e., the A2780 cell line, by the MTT assay. The representative dose-response profiles after 72 h of treatment are shown in Figure S3 together with the corresponding IC₅₀ values. These results point out that conjugation to HuHf produces a huge increase in the action of AuTM toward A2780 cancer cells, amounting to a factor of ~110X (0.43 vs 47 μM), calculated on a molar basis. These IC₅₀ concentrations can halve cell growth, although all cells in the samples result alive when tested via the Trypan Blue assay (Figure S4). To exclude any possible interference due to gold(I) binding to the free Cys34 of human serum albumin (HSA), which is present in the growth media at a concentration ∼60 μM, we investigated by ESI MS the interaction of AuTM-derived gold species with HSA (Figure S5A). Analogously to the case of AF,¹⁰ HSA with its single Cys does not compete with HuHf; additionally, no binding to HSA is observed even when incubating HSA with free AuTM up to a HSA : AuTM ratio of 1 : 5, which is larger than that used for cell treatment (Figure S5B).

To support this hypothesis and demonstrate that the association with ferritin may make AuTM more selective for cancer cells than for healthy cells, we have comparatively measured the level of transferrin receptor expression in A2780 cells and in Human Embryonic Kidney (HEK) cells, the latter taken as a model of non-cancerous cell lines. Consistently with literature data,^{18, 31, 32} TfR1 was found to be expressed in significantly higher levels in A2780 cells (Figure 6A). Using ferritin labeled with the Alexa Fluor 488 dye, we could observe the more efficient uptake of the nanocage by A2780 cells with respect to HEK cells (Figure 6B). In line with these results, the IC₅₀ at 72 h for HEK cells treated with HuHf@AuTM is >4 μM (on a ferritin molar basis), i.e., at least 10 times larger than the corresponding value for the A2780 cells.

Figure 7 shows the MTT time course (24, 48 and 72 h) performed upon treating A2780 cells with AuTM and HuHf@AuTM with the respective 72 h-IC₅₀ values, whereas Figure 7B compares the cellular gold uptake measured through ICP-OES determinations, at the same times of treatment and equal compound concentrations as for the MTT time course. Gold concentrations were determined on mineralized cellular pellets derived from viable cells (see Supporting Information for the detailed procedure). It is evident that the treatment with the HuHf@AuTM adduct resulted in a much higher cellular uptake of gold compared to the free drug (by a factor ~7.5 at 24 h, which decreased down to ∼3.0 at 48 h). The more efficient gold uptake might explain the differences in IC₅₀. These data strongly suggest that ferritin acts as an excellent nanocarrier for AuTM, greatly enhancing the effect of the free drug.

Afterward, to gain deeper insight into how the treatment with free AuTM and HuHf@AuTM affects cellular processes, untargeted ¹H NMR metabolomic analysis was performed both on cell lysates and growth media of A2780 cells treated with AuTM or with HuHf@AuTM at two different time points, 24 h and 48 h, using the respective IC₅₀ concentrations at 72 h (equal toxicity). Representative ¹H NMR spectra of cell lysates and growth media of untreated (control) cells and treated A2780 cells are provided in Figure S6 and Figure S7. The assigned metabolites are listed in Table S4. The pairwise comparisons between A2780 control cells and AuTM- or HuHf@AuTM-treated cells revealed significant alterations in response to these treatments, both in the media and in the lysates (Figure S8). The time dependence of the significant changes is depicted by the box plots of Figures 8 and 9, assembling changes as a function of glycolytic and TCA pathways, respectively.

The most relevant changes induced by AuTM and HuHf@AuTM treatment in A2780 cancer cells based on our NMR metabolomics data are summarised below:

To validate the conclusions drawn from the NMR metabolomics analysis, we carried out additional biochemical tests. In a first experiment we evaluated whether AuTM treatment may affect oxygen consumption as previously observed for AF and the related compounds gold(I) monocarbene and gold(I) dicarbene.³⁴ The results (Figure S10) indicated that AuTM did not affect oxygen consumption at all. Then, to confirm the alterations observed in the TCA cycle metabolites, succinate dehydrogenase (SDHA) and citrate synthase (CS) protein levels were measured in treated cells versus controls (Figure 9). A significant decrease in SDHA is observed for both treatments at both time points; this behaviour explains the observed increase in succinate (substrate) and decrease in fumarate (product) levels. A progressive decrease of CS is observed over time for both treatments, which is consistent with a decrease in citrate; Acetyl-CoA is below the detection limit of ¹H NMR.

Therefore, by integrating untargeted NMR and targeted biochemical tests, we obtained a fully consistent demonstration that AuTM and HuHf@AuTM induce substantially similar alterations in the metabolome. In turn, the latter differ profoundly from those previously reported for AF³³ and its ferritin adduct¹⁰ as well as for the gold(I) monocarbene and gold(I) dicarbene compounds.³⁴

Finally, we tested the activity of AuTM and HuHf@TM in a cellular model of A2780 cells resistant to cisplatin (named A2780cis, Figure S11A and B). Similarly to what observed for cisplatin sensitive A2780 cells, HuHf@AuTM has a much stronger effect than free AuTM, of the order of 60X, calculated on a molar basis. Both AuTM and HuHf@AuTM appear to overcome cisplatin resistance. Consistently, the analysis based on metabolomics shows that AuTM and its bioconjugate affect largely different pathways (Figure S11C) than cisplatin.