Interactions of Boron Clusters and their Derivatives with Serum Albumin

Boron clusters are polyhedral boron hydrides with unique properties, and they are becoming increasingly widely used in biology and medicine, including for boron neutron capture therapy (BNCT) of cancers and in the design of novel bioactive molecules and potential drugs. Among boron cluster types, icosahedral boranes, carboranes, and metallacarboranes are particularly interesting, and there is a need for basic studies on their interaction with biologically important molecules, such as proteins. Herein, we report studies on the interaction of selected boron clusters and their derivatives with serum albumin, the most abundant protein in mammalian blood. The interaction of boron clusters with albumin was examined by fluorescence quenching, circular dichroism, dynamic and static light scattering measurements and MALDI-TOF mass spectrometry. Our results showed that metallacarboranes have the strongest interaction with albumin among the tested clusters. The observed strength of boron cluster interactions with albumin decreases in order: metallacarboranes [M(C2B9H11)2]− > carboranes (C2B10H12) >> dodecaborate anion [B12H12]2−. Metallacarboranes first specifically interact with the binding cavity of albumin and then, with increasing compound concentrations, interact non-specifically with the protein surface. These findings can be of importance and are useful in the development of new bioactive compounds that contain boron clusters.

. NMR studies showed the formation of complexes with high stoichiometry caused by nonspecific binding of the metallacarborane on surface of HSA 15 . The interaction between [CoD] − and HSA is mainly driven by hydrophobic forces because the strength of this interaction is correlated with the lipophilicity (log P OW ) of the metallacarborane derivatives 16 .
[CoD] − also binds to the hydrophobic pocket of HIV protease and acts as a potent, specific and selective inhibitor of that enzyme 17,18 .
Dodecaborate is an icosahedral, dianionic, highly water soluble boron cluster. Although hydrophilic, dodecaborate forms strong inclusion complexes with γ-cyclodextrin 33 . The interaction with the hydrophobic cavity of cyclodextrin is driven by a chaotropic effect and not by hydrophobic forces. Derivatives of dodecaborate interact with lipid membranes 34 . Due to its high boron content, dodecaborate is commonly used as a boron carrier for BNCT. 10 B-sodium-mercaptoundecahydrododecaborate (BSH) 35 and 10 B-L-borophenylalanine 36 are used as boron carriers in BNCT 35 .
Almost every drug, theranostic or diagnostic agent interacts with albumin, the most abundant protein in mammalian blood. Serum albumin (SA) possesses binding ability for a broad spectrum of compounds, either endogenous and exogenous, due to the numerous binding regions and diverse specificity 37 . Bovine serum albumin (BSA) has three primary binding sites for long-chain fatty acids (one site in domain I and two sites in domain III) and two secondary sites [38][39][40] . With respect to exogenous compounds, BSA has two high affinity binding sites for drugs, known as the Sudlow's sites I and II, which are located in domains II and III 41 . Binding with serum proteins influences the pharmacokinetics and pharmacodynamics of these compounds 42,43 . Hence, investigation of the binding affinity of boron clusters to SA may provide useful information for designing novel boron-containing, biologically active molecules.
To the best of our knowledge, except for the use of HSA as a solubilizing agent for metallacarboranes 14,15 , there are no studies describing the interactions of boron clusters (and molecules containing these compounds) with serum proteins. The aim of this research is to address this gap and to characterize interactions between bovine serum albumin and various boron clusters. Boron clusters of three different types were selected for the study due to differences in their physicochemical properties and applications in medicinal chemistry.
Herein, the interactions of boron clusters and hybrid organic-boron cluster conjugates with BSA were investigated with fluorescence, absorption and circular dichroism spectroscopy as well as dynamic and static light scattering and mass spectrometry techniques. The pattern of the interactions with BSA and an effect of the boron clusters type on binding to the BSA molecule were established.

Results and Discussion
Fluorescence quenching mechanism. BSA, because of its structural homology to human serum albumin (HSA), is often used to study interactions with drugs 37 . No major differences between HSA and BSA have been observed in terms of the binding constant or binding mode 44,45 . The intrinsic fluorescence of BSA comes from tryptophan (Trp), tyrosine (Tyr) and phenylalanine (Phe) residues. However, the Trp residue has the strongest fluorescence intensity and is the most sensitive to changes in the microenvironment; therefore, it is indicative of the protein conformational alterations with binding. BSA has two Trp residues (Trp134 and Trp212), which are located within hydrophobic pockets in the first and second domains of BSA molecule 46 . Upon binding of a ligand in these pockets, the Trp environment changes, quenching the fluorescence.
In the first step, we measured the fluorescence intensity of BSA in the presence of boron clusters of three different types, metallacarborane (1); para-carborane derivative (7) and dodecaborate (11) (Fig. 1). The BSA: boron clusters molar ratio was 1 : 1. The parameter degree of quenching was calculated (Table S1). Based on these results, we decided to perform further measurements of the fluorescence intensity of BSA in the presence of metallacarboranes (compounds 1 to 6) with a maximum BSA : boron cluster molar ratio of 1 : 1 and to increase the ratio of the other boron clusters (7 to 11) to 1 : 10 ( Fig. 2, Fig. S1).
The fluorescence quenching mechanism can be divided into the following two mechanisms: dynamic quenching, caused by collisional encounters of fluorophores and quenchers, or static quenching, caused by ground-state complex formation between fluorophores and quenchers 47 . Dynamic and static quenching can be distinguished by their dependence on temperature and excited-state lifetime. If the value of the quenching constant (K SV ) decreases with the increasing temperature, the mechanism is usually static because the higher temperature is likely to decrease the stability of the complex. If K SV increases with increasing temperature, the mechanism is likely dynamic because the higher temperature results in a larger diffusion coefficient and promotes electron transfer. The quenching constant (K SV ) was calculated using the Stern-Volmer equation (1) 47 : where F 0 and F are the steady-state fluorescence intensities in the absence and presence of quencher, respectively, and [Q] denotes the initial concentration of quencher. k q is the quenching rate constant of BSA and τ 0 is the fluorescence lifetime of BSA in the excited state without any quencher, and its value is 6×10 −9 s 48 . K SV can be calculated from the slope of the Stern-Volmer equation (Fig. S2).
To determine the mechanism of fluorescence quenching in the case of boron clusters and BSA, the measurements of fluorescence quenching were performed at three temperatures (25, 31 and 37 °C). We observed differences among different types of boron clusters. Hydrophobic metallacarboranes (1 to 6) were characterized by the highest value of K SV on the order of 10 4 to 10 5 (Table S2). For doubly negatively charged dodecaborate 11, the fluorescence quenching was not observed. For neutral para-carborane derivatives (7 to 9), nonlinear dependence of quencher concentration and fluorescence quenching were observed, except for the conjugate of carborane and adenosine, where the correlation was linear. Therefore, the K SV value could not be calculated for the last two boron cluster types. For metallacarboranes (1 to 6), we observed an increase of the K SV value with increasing temperature, which indicates a dynamic mechanism of quenching. However, the values of the quenching rate constants (k q ) are 2-to 3-fold higher than the maximum scatter collision quenching constant for various quenchers with biopolymers (10 10 L mol −1 s −1 ) 49 , suggesting that static quenching also occurs.

Binding constant and binding modes.
For the boron cluster -BSA interaction, the binding constant (K b ) and number of binding sites (n) of the complexes can be calculated by the equation (2), under assumption that BSA has independent binding sites 47 .
The K b and n values can be calculated by the intercept and slope of the regression curve (Fig. S3). The estimated K b values were on the order of 10 5 M −1 (Table S3) for metallacarboranes (1 to 6), which is similar to the values of the binding constant for small organic drugs that have a strong binding force with BSA, such as aspirin 41 . The values of n were approximately equal to 1 in the studied temperature range, indicating the existence of a single binding site of boron clusters on BSA.
In the binding process for drugs with proteins, several binding forces can be engaged, including hydrogen bonding interaction, van der Waals forces, electrostatic interaction and hydrophobic interaction. The distinction between those forces can be made based on the signs and magnitudes of the thermodynamic parameters, such as the Gibbs free energy change (ΔG 0 ), enthalpy change (ΔH 0 ) and entropy change (ΔS 0 ). The thermodynamic parameters can be calculated by the equations (3) and (4).
where R is the gas constant and K b is the binding constant at corresponding temperatures (Fig. S4). Based on the thermodynamic view point 50   and/or hydrogen bonding interaction, and (III) ΔH 0 close to 0 and positive value of ΔS 0 for the electrostatic interaction.
The positive values of both ΔH 0 and ΔS 0 for 1 to 6 (Table S3) suggest that the interaction mode for metallacarboranes binding on BSA is mainly a hydrophobic interaction. This result is in agreement with studies by Rak et al. in which the strength of the interaction with BSA was correlated with the lipophilicity of metallacarborane derivatives, suggesting the hydrophobic nature of the interaction 16 . Moreover, the negative value of ΔG 0 suggests that binding of the boron clusters with BSA is a spontaneous process.
Rak et al. reported a non-specific interaction of (1) derivatives with HSA for complexes with high stoichiometry 15 . Our studies supplement those observations with studies of complexes that have low stoichiometry with a maximum ratio of metallacarborane : BSA set to 1 : 1. In contrast to the previous report, we have found that under low stoichiometry conditions, metallacarboranes have a specific interaction with BSA by binding in the hydrophobic cavity of the protein. These observations lead to hypothesis that at low stoichiometry, metallacarboranes interact specifically with the hydrophobic cavity of albumin, while at high stoichiometry, they also interact non-specifically with the protein surface.
For further studies, we have chosen [CoD] − (1) and [B 12 H 12 ] 2− (11) as compounds with the strongest and weakest interactions with BSA, respectively. In that way, we could study the influence of the boron cluster types on the interaction with BSA in two extreme cases. Incorporation of [CoD] − into the structure of bioactive molecule can significantly affect its interactions with serum albumin. By contrast, [B 12 H 12 ] 2− does not seem to affect the affinity of the conjugate towards albumin. Furthermore, in contrast to the other two types of boron clusters, the number of studies on the interaction of [B 12 H 12 ] 2− with proteins is very limited. Considering that derivatives of this boron cluster are used as boron carrier in BNCT, there is a gap in our knowledge that needs to be addressed.

Conformational changes of BSA induced by boron clusters.
The CD spectrum of native BSA (25 °C, NaHCO 3 10 mM) exhibited two negative minima at 208 and 222 nm and a maximum at 192 nm, which is in good agreement with previous studies 51 and characteristic for proteins with high helical content. The CD spectra of native BSA in the presence of an increasing concentration of selected boron cluster (from 0 to 1000 molar excess) are represented in Fig. 3. The influence of [B 12 H 12 ] 2− on the BSA structure revealed in the far UV CD spectra is negligible. No essential conformational changes were observed for BSA in the presence of [B 12 H 12 ] 2− (0 to 1000 molar excess), indicating that the presence of the boron cluster cage did not significantly alter the tertiary structure of BSA (Fig. 3A).
The obtained data show that the portion of α-helices in the secondary structure of BSA is equal to 59% and increasing the concentration of [CoD] − in the BSA environment leads to a slight decrease in the number of α-helices (up to 54% for 512 molar excess of the boron cluster - Fig. 3B). The observed changes in the content of α-helices caused by [CoD] − are noticeable, but they seem to be negligible compared with changes in the BSA α-helix content caused by ionic surfactants such as SDS 51,52 . This observation is of interest because [CoD] − may behave like a classical ionic surfactant in selected cases, and it is able to form complexes with polyelectrolytes 12 . Features of near-UV CD spectra are more interesting. First, the presence of [B 12 H 12 ] 2− (0 to 100 molar excess) does not disturb the CD spectrum (Fig. 3C). In the case of [CoD] − , the situation is radically different. The presence of a metallacarborane cluster (0-8 molar excess) leads to formation of two new CD signals (Fig. 3D). Two positive Cotton effects at 350 nm and 309 nm are directly connected with complex formation between [CoD] − and the hydrophobic pocket of BSA. Slight changes in the CD spectrum below 300 nm suggest that this process does not significantly alter the tertiary structure of BSA. In addition, the formation of the [CoD] − : BSA complex is shown in the absorption spectrum (Fig. 3F) where increasing the concentration of [CoD] − causes the formation of two new absorption spectra (with a maximum at 350 nm and 302 nm) as well as the simultaneous decrease in the characteristic band for BSA at 278 nm and an isosbestic point at 291 nm.
Hydrodynamic changes of BSA induced by boron clusters. The weak protein-protein interactions, i.e., non-specific interactions, were studied using a combination of DLS and SLS techniques. Here we used two parameters to describe the BSA-BSA interactions in the presence of selected boron clusters, i.e., diffusion interaction parameter (k D ) and second virial coefficient (B 22 ), both of which are first-order coefficients of protein interactions. The diffusion interaction parameter (k D ) is a useful factor to study weak protein interactions and is connected to specific biophysical properties, such as the aggregation propensity. The diffusion coefficient (D) was measured as a function of the BSA concentration in the presence of a defined boron cluster using the DLS protocol. The value of k D (mL g −1 ) was determined from the slope of a plot of D vs. the BSA concentration, where D 0 is the extrapolated intercept (single particle diffusivity) and the slope is D 0 kD (equation 5).
The second virial coefficient (B 22 ), is another parameter for studying weak protein interactions. Positive B 22 is interpreted as indicative of the presence of weak, net-repulsive forces between protein molecules, while negative B 22 is considered to reflect net-attractive forces. The normalized Rayleigh ratio R θ from the SLS measurement is related to the properties of the protein solution (equation 6).  53 . The presence of positive slopes in both DLS and SLS data in the case of two distinct boron clusters, together with the linear behavior of both data, support that the potential contributions of these boron clusters to protein aggregation are negligible (Fig. 4). In addition, in the case of  (Fig. 4A), but [CoD] − exerted a pronounced influence on the hydrodynamic size of BSA, as observed in Fig. 4B for a [CoD] − : BSA molar ratio higher than 10 for which the hydrodynamic diameter of BSA (or [CoD] − : BSA complex) slightly increased up to 9.5 nm. Next, despite the increase in the concentration of metallacarborane, we observed a fixed hydrodynamic size. Subsequently, for molar ratios higher than 100, the hydrodynamic size of BSA again starts to increase, suggesting the appearance of second generation complexes. Considering the shape and dimensions of [CoD] − (0.6×1.1 nm peanut shape), we can interpret these data as indicative of a non-specific interaction of [CoD] − with the BSA surface, which gives rise to a first interaction layer followed by an increasing [CoD] − : BSA molar ratio over 100; then, metallacarborane clusters begin to form the second interaction layer. Additional information on the behavior of BSA in the presence of boron clusters was obtained by DLS in the heat-treating experiment (Fig. S5). This method allows for registration of changes in the hydrodynamic size of the protein to detect aggregation. According to the literature reports 54,55 , the hydrodynamic size of BSA should significantly increase for temperatures higher than 60 °C. For BSA in the absence of the boron clusters as well as for BSA in the presence of [B 12 H 12 ] 2− , our results are in good agreement with these reports. Sodium dodecaborate did not influence the hydrodynamic behavior of BSA during thermal treatment (Fig. S5C). A different situation was observed in the case of [CoD] − . Metallacarborane exerted a profound influence on BSA during heating when the temperature ranged from 25 °C to 85 °C, and this process depends on the [CoD] − concentration (Fig. S5D).
[CoD] − can limit thermal-induced BSA aggregation and the resulting aggregates are much smaller than native BSA, which is also noticeable after cooling to room temperature. In addition, [CoD] − influences the aggregation kinetics at a constant temperature of 60 °C (Fig. S5F). Selected cases have inhibition of aggregation at that temperature.

MALDI-MS.
We also demonstrated the interactions of boron clusters with BSA using the MALDI-MS technique. A molecular mass measurement was made as a function of the boron cluster concentration and the stoichiometry of the protein : boron cluster complex was calculated from these data. The results of the formation of non-covalent binding complexes between [CoD] − or [B 12 H 12 ] 2− and BSA are summarized in Fig. 5. Both types of boron clusters can form non-covalent complexes with BSA. Moreover, the measured molecular weight increased linearly with the boron cluster concentration, allowing for the calculation of the stoichiometry of complexes formed. It is noteworthy that the measurement was made in acidic MALDI matrix, in the presence of organic solvent and that the conditions did not cause dissociation of the complexes. Changes in the molecular weight are even significant at a [CoD] − : BSA molar ratio of 2 : 1. This observation suggests that complex formation between [CoD] − and BSA is very efficient and the resulting complexes very stable. In contrast, the interaction of [B 12 H 12 ] 2− with BSA is much weaker (cf. experimental) and only visible at higher molar ratios.

Conclusions
The interaction between BSA and three types of boron clusters (metallacarboranes, para-carboranes and dodecaborates) and their derivatives was studied using measurements of fluorescence quenching of BSA in increasing  Carboranes (7 to 10) also interact with BSA; however, the relationship between the boron cluster concentration and fluorescence quenching of BSA does not follow a linear correlation, except for the conjugates of carborane and adenosine (10). The results for [B 12 H 12 ] 2− show no interaction of this boron cluster with BSA. The measurements of fluorescence quenching were performed at a low stoichiometry of BSA and studied compounds, which was a maximum 1 : 1 for metallacarboranes and their derivatives (1 to 6) and 1 : 10 for other boron cluster compounds (7 to 11). Further studies with [CoD] − and [B 12 H 12 ] 2− , using CD, DLS and SLS measurements as well as the MALDI-MS technique, were performed at a high stoichiometry up to 1: 1000. The results of these studies showed the non-specific binding of selected boron clusters with the BSA surface. Interactions of BSA with [CoD] − are much stronger and have a bigger influence on the conformation of the protein than with [B 12 H 12 ] 2− . [CoD] − specifically binds to the hydrophobic cavity of BSA at low stoichiometry and non-specifically binds with the surface of the protein at high stoichiometry. [B 12 H 12 ] 2− interacts very weakly with BSA and only in the non-specific way at a high stoichiometry.  (6) were obtained according to previously published methods 60 ; detailed experimental procedures will be described elsewhere. 1-(3-Bromopropanyl)-1,12-dicarba-closo-dodecaborane (7) and 1-(3-aminopropanyl)-1, 12-dicarba-closo-dodecaborane (9) were synthesized following the method described earlier for the analogous derivatives of 1,7-dicarba-closo-dodecaborane 61 . 1-Carboxyl-1,12-dicarba-closo-dodecaborane (8) was obtained as described 62  Sample preparation. The stock solutions of the compounds 1-11 (10 mg mL −1 ) were prepared by dissolving the exact amounts in DMSO (compounds 2-10) or 0.1 M NaHCO 3 solution (compounds 1 and 11). The stock solution of BSA (0.67 mM) was prepared using dialysis with 0.1 M sodium bicarbonate (pH 8.4), and its concentration was determined by the measurement of the absorbance at λ = 278 nm using ε = 44300 M −1 cm −1 64 . The BSA samples with boron clusters for fluorescence, UV-Vis, CD and light scattering analysis were prepared as follows: the stock solution of selected boron clusters was added to the BSA solutions in the desired molar ratio and measurements were taken immediately after mixing.

Materials and equipment. Cs
Experimental procedure UV-Vis measurements. In all measured samples, BSA was diluted with 0.10 M sodium bicarbonate (pH 8.4) to a final concentration 10.5 × 10 −6 M. Boron clusters 1 to 6 were added to the samples so the final concentration was 0 to 10.5 × 10 −6 M. Boron clusters 7 to 11 were added to the sample so the final concentration was 0 to 10.5 × 10 −5 M. In all samples, the concentration of DMSO was 2%. Absorption spectra were recorded from 250 to 500 nm. Each spectrum was corrected by the corresponding buffer blank.
Fluorescence measurements. For fluorescence measurements, the samples were prepared in the same way as for UV-Vis measurements. Fluorescence emission spectra were recorded from 300 to 500 nm at three different temperatures (25, 31 and 37 °C with 3 min equilibrium time at each measurement temperature). The excitation wavelength was set at 280 nm. The data integration time was set to 1 s and HT to 600 V. An appropriate buffer blank spectrum was subtracted from the measured spectra for fluorescence background correction. The fluorescence intensity was corrected using the following equation: where F corr and F obs are the corrected and observed fluorescence intensities, respectively and A ex and A em are the absorbances at the excitation and emission wavelengths, respectively. The emission wavelength was 355 nm at the maximum of the fluorescence intensity spectrum. All fluorescence intensities used for the calculations were corrected in this work.
Circular dichroism measurements. The BSA concentration was 1.5 μM for far UV measurements and 150 μM for near UV, and the final concentration of NaHCO 3 was 10 mM for all CD measurements. Six spectra (recorded with a data pitch of 0.2 nm, band width of 1 nm, and data integration time of 2 sec at 50 nm min −1 ) were averaged for each sample. Each measurement was subtracted by NaHCO 3 (10 mM) supplemented with a suitable amount of selected boron cluster. Therefore, the emerging spectrum is due to the BSA contribution alone. The CD signals were converted to the mean residue molar ellipticities using the mean residue weight of 114.2. The fractional contents of α-helices were calculated from far-UV CD spectra by the Dichroweb platform (CDSSTR with dataset 6). Data Availability. All data generated or analysed during this study are included in this published article (and its Supplementary Information files).