lable at ScienceDirect

Fluid Phase Equilibria 422 (2016) 18e31

Contents lists avai

Fluid Phase Equilibria

journal homepage: www.elsevier .com/locate/fluid

Structure and binding thermodynamics of viologen-phosphorousdendrimers to human serum albumin: A combined computational/experimental investigation

Erik Laurini, Domenico Marson, Paola Posocco, Maurizio Fermeglia, Sabrina Pricl*

Molecular Simulation Engineering (MOSE) Laboratory, Department of Engineering and Architecture (DEA), University of Trieste, Piazzale Europa 1, 34127Trieste, Italy

a r t i c l e i n f o

Article history:Received 30 October 2015Accepted 9 February 2016Available online 15 February 2016

Keywords:Human serum albuminDendrimersFree energy of bindingMolecular simulationsExperimental validation

* Corresponding author.E-mail addresses: erik.laurini@di3.units.it (E. Lau

units.it (D. Marson), paola.posocco@di3.units.it (P. Pdi3.units.it (M. Fermeglia), sabrina.pricl@di3.units.it (

http://dx.doi.org/10.1016/j.fluid.2016.02.0140378-3812/© 2016 Elsevier B.V. All rights reserved.

a b s t r a c t

Low-generation viologen-phosphorous dendrimers (VPDs) can be exploited as novel therapeutic agents,since they efficiently inhibit aggregation of amyloid-b into fibrils and are active against several strains ofmicroorganisms. Human serum albumin (HSA), the most abundant plasma protein, is playing anincreasing role as drug carrier in the clinical setting. Therefore, with the aim of exploiting HSA as apotential carrier for VPDs, in this work we performed a preliminary investigation of the interaction of sixdifferent VPDs 1e6 with HSA using a combined computational/experimental approach. First, differentmodeling techniques were employed to i) determine the dendrimer binding site on the HSA surface; ii)derive the free energy change DGb involved in each dendrimer/HSA complex formation; iii) analyze indetails all molecular determinants contributing to DGb, and iv) evaluate the eventual HSA structuralvariations induced by dendrimer binding. All modeling predictions were next validated using a series ofexperimental techniques, including isothermal titration calorimetry (ITC), circular dichroism (CD), andfluorescence quenching and decay. In aggregate, the results from this study allowed us to rank the af-finity of the different viologen-phosphorous dendrimers 1e6 towards HSA and to formulate a molecular-based rationale for the differential binding thermodynamics of the resulting dendrimer/HSA complexes.According to our data, HSA can successfully and selectively bind VPDs 1e6, dendrimer 4 being the bestcargo for this endogenous protein nanocarrier.

© 2016 Elsevier B.V. All rights reserved.

1. Introduction

Human serum albumin (HSA) is the most abundant plasmaprotein (30e50 g/L human serum). With a molecular weight of66.5 kDa and an average half-life of 19 days, this protein is syn-thesized in the liver (10e15 g/day). HSA is best known for itsextraordinary ligand binding capacity, providing a depot for a widevariety of compounds that may thus be available in quantities wellbeyond their solubility in plasma.

HSA abundance makes it an important factor in the pharma-cokinetic behavior of many natural metabolites (i.e., fatty acids,hormones, bilirubin, tryptophan, steroids, and metal ions) and of aplethora of therapeutic agents, affecting their efficacy and rate of

rini), domenico.marson@dia.osocco), maurizio.fermeglia@S. Pricl).

delivery. This aspect has stimulated many efforts to understandwhat determined HSA/drug binding. Binding to HSA is also at theroot of the development of contrast agents, for magnetic resonanceimaging (MRI) endowed with high intravascular retention (bloodpool agents), such as those used in the clinical practice for thevisualization of vascular structures (magnetic resonance angiog-raphy) and for detecting regions with abnormal vascular perme-ability. In other cases, HSA holds some ligands in a strainedorientation, providing their metabolic modifications, and renderspotential toxins harmless by transporting them to disposal sites[1e3]. After acute hemolysis (e.g., after trauma or post-ischemicreperfusion), HSA binds heme that is released in the bloodstream. Heme is then gradually transferred from HSA to humanhemopexin, which allows its receptor-mediated re-uptake byparenchymal liver cells [4]. HSA also plays many other fundamentalbiological roles; among these, this protein i) is the essential regu-lator of blood osmotic pressure; ii) accounts for most of the anti-oxidant capacity of human serum, either directly or by binding and

E. Laurini et al. / Fluid Phase Equilibria 422 (2016) 18e31 19

carrying radical scavengers, or by sequestering transition metalions endowed with pro-oxidant activity; and iii) acts as a NO re-pository and carrier, leading to covalent modification(s) of (macro)molecules [1].

Today, HSA is playing an increasing role as a drug carrier in theclinical setting, via its coupling to low molecular weight drugs, orits conjugation with bioactive protein/macromolecules, or theencapsulation of drugs into albumin nanoparticles [5]. The under-lying strategy is that binding of a therapeutic agent e eithercovalently or physically - to human serum albumin might increasedrug stability, half-life, biodistribution, and bio-barrier (e.g.,bloodebrain barrier or BBB) permeability [3,5]. The peculiarchemico-physical characteristics of HSA (which include, amongothers, stability in a wide pH range (4e9), high solubility, andexceptional resistance to denaturation), coupled with its prefer-ential uptake in tumor and inflamed tissues, its ready availability,its biodegradability, and its lack of toxicity an immunogenicitymake it an ideal nanocarrier candidate for drug delivery.

Dendrimers undoubtedly constitute one of the most importantcontributions of chemistry to nanoscience [6e9]. Their branched,perfectly defined, layered, and three-dimensional architecture hasa size varying generally from 1 to less than 20 nm, depending onthe generation (i.e., the number of self-similar layers). Indepen-dently of the chemical nature, a dendrimer structure is composedof branches constituted of repeating units, emanating from a cen-tral core. The extremities of the branches (also called terminalunits/groups) are generally reactive and can be functionalized atwill, depending on the properties that are targeted [10].

There are several main families of dendrimers, but new types ofdendritic structures are constantly being synthesized. One novelclass of dendritic compounds is viologen-phosphorus dendrimers(VPDs). Phosphorus-containing dendrimers [11,12] are consideredto be particularly important for biomedical research becausephosphorus is essential for all terrestrial forms of life. The biologicalproperties of these compounds have been recently reviewed indetail, including their positive effect on the growth of neuronalcells, monocytes and natural killer cells, their anti-prion properties,their use as delivery platforms for ocular drugs and transfection andimaging agents, and their potential as highly sensitive biosensors[13]. Used for a long time as unselective pesticides (aka Paraquat),viologen derivatives (4,40-bipyridinium salts) can also exert severeif not lethal poisoning in man [14]. The mechanism of viologentoxicity involves the induction of generation of superoxide anionsand other reactive oxygen species (ROS), which ultimately resultinto cell and tissue damage [15]. Moreover, there is increasing ev-idence supporting the association of chronic Paraquat exposure andParkinson's disease [16].

Interesting, however, concomitant studies have shown that theincorporation of viologen units as part of a dendrimeric structurecan yield new molecules with surprisingly beneficial biologicalactivities [17e19]. Thus, several cationic viologen-based den-drimers were tested for and found endowed with encouragingantiviral activity against human immunodeficiency virus (HIVeI),herpes simplex virus (HSV), vesicular stomatitis, Punta Toro virus,Sindbis virus, Rheovirus and Respiratory Syncytial virus (RSV) [20].

Recently, the group of Caminade and Majoral has synthesized anovel series of dendrimers featuring both viologen units andphosphorous groups in their structure [21], which differ in natureof their core (tri-vs. hexa-functionalized), generation (G0 vs. G1),number of viologen units, number and type of terminal groups(-CHO vs. PO(Et)2). Experiments in the field of neurodegenerativedisorders (e.g., transmissible spongiform encephalopathies, Alz-heimer and Parkinson's diseases) were carried out with theseviologen-phosphorous dendrimers (VPDs) [22]. Quite encourag-ingly, all these VPDs, and specifically those bearing the

phosphonate (-PO(OEt)2) terminal group, effectively inhibitedacetylcholinesterase and butyrylcholinesterase, two enzymes thatare implicated in Alzheimer's disease as they co-localize with am-yloid-b (Ab) peptide plaques and accelerate the assembly of Ab intofibrils [23]. More, these compounds also bind and influence theactivity of a-synuclein, another protein involved in the sameneurodegenerative pathology, thereby inhibiting its fibrillation[24]. Another series of biological tests with the same VPDs con-cerned their hemotoxicity, cytotoxicity, and antibacterial activity[21]. It was shown that those compounds of generation 1 (i.e.,bearing the highest number of positive charges) induced thehighest hemotoxicity (i.e., hemolysis); contextually, all VPDs werefound to be non-toxic towards B14 (healthy) cells but, at the sametime, very toxic towards cancerous N2a cells. In addition, most ofthe tested VPDs exhibited good antimicrobial properties towardsS. aureus strain (Gram-positive bacteria) while the most-chargeddendrimers of the series also limited the growth of Gram-negative bacterial strain (E. coli and P. vulgaris).

Notwithstanding the potential perspective of biomedical ap-plications of viologen-phosphorus dendrimers outlined above,quite surprisingly no studies beyond basic biological properties(mainly cytotoxicity assays) of VPDs have been published so far.Since the possible exploitation of viologen-phosphorus dendrimersas nanomedicines per se requires a deeper knowledge of theirbiological behavior, in this work we decided to investigate theinteraction of VPDs with human serum albumin. As outlined above,VPDs are able to interact with several, different human proteins;thus, we reasoned that cationic viologen-phosphorous dendrimerscould exhibit a natural affinity for HSA, the most abundant andhighly negatively charged plasma protein. Should this hypothesisbe verified, the drug carrier native propensity of HSA could furtherbe exploited as generate a novel endogenous nanocarrier/cargosystem for the VPDs. Accordingly, we performed a preliminaryinvestigation of the interaction of six different VPDs with HSA usinga combined computational/experimental approach. First, differentmodeling techniqueswere employed to i) determine the dendrimerbinding site on the HSA surface; ii) derive the free energy changeDGb involved in each dendrimer/HSA complex formation; iii)analyze in details all molecular determinants contributing to DGb,and iv) evaluate the eventual HSA structural variations induced bydendrimer binding. All modeling predictions were next validatedusing a series of experimental techniques, including isothermaltitration calorimetry (ITC), circular dichroism (CD), and fluores-cence quenching and decay. In aggregate, the results from thisstudy allowed us to rank the affinity of the different viologen-phosphorous dendrimers towards HSA and to formulate amolecular-based rationale for the differential binding thermody-namics of the resulting dendrimer/HSA complexes.

2. Materials and methods

All chemicals were purchased from Sigma Aldrich and AcrosOrganics, and used without further purification. Human serum al-bumin (fatty acid free) was also purchased from Sigma and used assupplied. The dendrimer cores trihydrazidophosphine sulfideSP(NMeNH2)3 and hexahydrazido cyclophosphazeneN3P3(NMeNH2)6 were prepared following literature indications[25]. 1H and 13C NMR spectra of the dendrimers were recorded atroom temperature with a 500 MHz Varian NMR spectrophotom-eter, using SiMe4 as internal standard for NMR chemical shifts. Thefollowing notation is adopted for NMR splitting patterns: s, singlet;d, doublet; t, triplet; q, quartet; m, multiplet. Fourier-transformedinfrared spectra (FT-IR) were recorded on neat samples using aPerkineElmer Spectrum RXI. Mass spectrometry (ESI-MS, positivemode) was carried out with a PerkineElmer PE-API I Spectrometer.

E. Laurini et al. / Fluid Phase Equilibria 422 (2016) 18e3120

2.1. Synthesis of viologen-phosphorous dendrimers

Six viologen-phosphorous dendrimers 1e6 with different endgroups and number of viologen units were synthesized and studiedin this work. Their chemical structures are shown in Fig. 1 whiletheir main characteristics are listed in Table 1. All dendrimers weresynthesized according to the procedure described by Ciepluch et al.[21]. The synthetic pathways and NMR characterization of alldendrimer molecules are reported in the Supporting Information.

2.2. Fluorescence quenching

All steady-state fluorescence spectra were recorded using aHitachi F-7000 fluorescence spectrophotometer (Hitachi, Japan)equipped with a thermostatic cell holder and a 1 cm path lengthcell cuvette. Fluorescence quenching experiments were carried outby fixing HSA concentration (PBS buffer, pH 7.4) to 5 � 10�6 mol/Lwhereas dendrimer concentration was varied in the range0e20 � 10�6 mol/L. Measurements were performed at 298, 303,and 310 K. An excitation wavelength of 290 nm was used andemission spectrawere recorded from 300 to 440 nm. The excitationand emission slits were set to 5 nm and 10 nm, respectively. Time-resolved fluorescence decays were determined by the time-correlated single-photon counting (TCSPC) spectrometer on aHoriba Jobin Yvon Fluoromax4 spectrophotometer, using a pico-second diode laser (nanoLED-07) as the light source at 295 nm andTBX-04 detector (both from IBH, U.K.). The observed fluorescencetransients were fitted using a nonlinear least-square fitting proce-dure [26].

2.3. Circular dichroism (CD) spectra

CD measurements of PBS buffered solutions of HSA in theabsence and presence of VPDs 1e6 were carried out on a JascoCorporation J-815 CD spectrophotometer (Jasco, USA) using aquartz cuvette of path length 1.0 cm at 1 nm pitch intervals. All CDspectra were recorded in the wavelength range 200e250 nm. Thespectrophotometer was continuously purged prior and during theexperiments with 99.9% nitrogen. Spectra were collected at scanspeed of 50 nm/min with response time of 1 s at 298 K. Eachspectrumwas baseline corrected and the final plot was taken as anaverage of four accumulated plots. Results were expressed as meanresidue ellipticity MRE, calculated according to the formula:MRE ¼ (MREobs/10) (MRM/LCp), where MREobs is the observedellipticity (mdeg), MRM is the mean residue molecular mass, L is

Fig. 1. Chemical structures of the violog

the optical path length (cm), and Cp is the protein concentration(mol/L). To calculate the composition of the secondary structure ofthe protein [27], the SELCON3 [28], CONTIN [29], and CDSSTRprograms [30] e as implemented in the CDPro software package -were used to analyze far-UV CD spectra. Final results were assumedwhen data generated from all programs showed convergence [31].

2.4. Isothermal titration calorimetry (ITC)

ITC experiments were conducted using a Nano ITC Technology(TA Instruments). Binding conditions were optimized for eachviologen-phosphorous dendrimer. All experiments were per-formed in PBS buffer at pH 7.4. All solutions and buffers used in theexperiments were degassed for 30 min at room temperature understirring at 600 rpm prior to experiment. Upon filling cell and sy-ringe, stirring was turned and the each system was allowed tothermally equilibrate for 1 h. For background correction, buffer (incell) was titrated with the injection of dendrimers 1e6 at the sameconcentration; background was then subtracted from final curves.The binding and thermodynamic parameters: constant of binding(Kb), the number of binding centers per one molecule (n), andenthalpy of binding (DHb) were computed from actual calorimetricdata by a non-linear fitting using the ITC software Origin. Using thefit output of DHb and Kb, the free energy of binding DGb and itsentropic component eTDSb were simply determined using stan-dard thermodynamic relationships (i.e., DGb ¼ �RT ln Kb andeTDSb ¼ DGb�DHb). Statistics were performed on the thermody-namic parameters with a desired confidence interval of 95%. Eachexperiment was repeated at least in triplicate, and the average ormultiple runs was used to obtain Kb.

2.5. Molecular simulations

2.5.1. Model building, refinement and simulation detailsAll simulations discussed in this workwere carried out using the

AMBER14 suite of programs [32] and performed with the GPUversion of pmemd (pmemd.cuda) in AMBER14 on the EURORA GPU-CPU supercomputer (CINECA, Bologna, Italy). The entire MDsimulation and data analysis procedure was optimized by inte-grating Amber14 in modeFRONTIER, a multidisciplinary and mul-tiobjective optimization and design environment [33]. The sixdendrimer models were built, parameterized and refined followinga consolidated procedure described in details in our previous works[34e41]. Briefly, the 3D structure of each dendrimer was built andgeometry-optimized using the Antechamber module of AMBER14

en-phosphorous dendrimers 1e6.

Table 1Main physico-chemical characteristics of viologen-phosphorous dendrimers 1e6.

Dendrimer Core Generation Charge Terminal group Number of terminal groups

1 S¼P(NMeeN¼)3 0 þ6 CHO 32 (N3P3) (NMeeN¼)6 0 þ12 CHO 63 S¼P(NMeeN¼)3 0 þ6 P(O) (OEt)2 34 (N3P3) (NMeeN¼)6 0 þ12 P(O) (OEt)2 65 S¼P(NMeeN¼)3 1 þ18 P(O) (OEt)2 66 (N3P3) (NMeeN¼)6 1 þ36 P(O) (OEt)2 12

E. Laurini et al. / Fluid Phase Equilibria 422 (2016) 18e31 21

consistently with General Amber Force Field (GAFF) [42]. Eventualmissing force field terms were derived from quantum mechanical(QM) calculations using GAMESS software [43,44] and the paramfitfeature of AMBER14. For QM calculations, the MP2/HF/6-31G levelof theory was used. Partial charges were obtained via the respprogram implemented in AMBER14.

The x-ray available crystal structure of ligand-free human serumalbumin (PDB code 1AO6.pdb) [45] was taken as starting point forprotein model preparation. The structures of each dendrimer andthat of the HSA were immersed in a box of TIP3P water molecules[46]. The dimension of each simulation box was chosen in order toensure a 1 nm solvation shell around each solute structure. Suitableamounts of Naþ and Cl� ions required to achieve solution neutralityand to realize a physiological ionic strength of 0.15Mwere added toeach system. The resulting hydrated structures were then subjectedto an initial Steepest Descent (SD)/Conjugated Gradient (CG)minimizationwith 5.0 kcal/(mol*A2) restraint on the solute (solventrelaxation), followed by another round of CGminimizationwithoutrestraints in order to eliminate all bad contacts between watermolecules and the dendrimer/HSA structure.

Next, each minimized structure was subjected to moleculardynamics (MD) simulations in the canonical ensemble (constantvolume/constant temperature, or NVT). During these 100 ps of MD,each systemwas gradually heated and relaxed to 300 K. The SHAKEalgorithm [47] was applied to all covalent bonds involvinghydrogen atoms. An integration time step of 2 fs was adoptedtogether with the Langevin thermostat for temperature regulation(collision frequency ¼ 2.0 ps�1) [48]. The final heating step wasfollowed by 50 ns of MD equilibration in the isochoric/isothermal(NPT) ensemble. Pressure control was exerted by coupling thesystem to a Berendsen barostat (pressure relaxation time 2 ps) [49].The particle Mesh Ewald (PME)method [50] was used to treat long-range electrostatic interactions under periodic conditions with adirect space cut-off of 10 Å2. A frame from each equilibrated MDtrajectory of the dendrimers and HSA was extracted to build thedifferent protein/dendrimer complex initial configuration.

For the construction of the dendrimer/HSA complex models, weresorted to Steered Molecular Dynamics (SMD) simulations[35,51e56]. Specifically, the dendrimer and HSA structuresextracted from the corresponding equilibrated MD simulationswere placed 60 Å away from each other in a solvated box. Next, thedendrimer was pulled close to its target protein by using a force of50 kcal/(mol*Å2) and a velocity of 5 Å/ns. The backbone atoms ofHSA were forced in their position by applying a weak restraint of0.5 kcal/(mol*Å2). This allowed avoiding substantial deformation ofthe protein during the dendrimer pulling process. Once the den-drimer reached the proximity of HSA (i.e., distance between thedendrimer and HSA center of mass approximately 12 Å), this re-straint was released and the both molecules were allowed to moveto reach the final complex configuration.

Each resulting dendrimer/HSA complex was again equilibratedfor 50 ns in the NPT ensemble and, starting from the last equili-brated frame, further 50 ns of simulation in an NVT ensemble wereperformed for structural and binding thermodynamic data

collection and analysis.

2.5.2. Structural analysisThe structural analysis of the HSA in complex with the den-

drimers was carried using the cpptraj program of AMBER14 and in-house developed python scripts. If not differently stated, all struc-tural data presented in this work represent values averaged overthe last 40 ns of the production runs, with MD trajectory snapshotstaken every 40 ps.

2.5.3. Free energy of binding calculationsThe free energy of binding DGb between each dendrimer and

HSA was derived by applying our thoroughly validated methodol-ogy [34e41] based on theMolecular Mechanics/Poisson BoltzmannSurface Area (MM/PBSA) approach [57]. This computational tech-nique employs snapshots taken from MD trajectories to estimatethe average interaction energies based on the solute molecularmechanics internal energy (DEMM) and solvation energy (DGsolv),this last obtained via PoissoneBoltzmann (PB) continuum solventcalculations. According to MM/PBSA, the overall binding energyDGb is given by the difference in energy between the HSA/den-drimer complex and the individual dendrimer and HSA molecules:

DGb ¼ DGcomplex � DGdendrimer � DGHSA (1)

where:

DGb ¼ DEMM þ DGsolv � TDSbind (2)

DEMM ¼ DEvdW þ DECoul (3)

DGsolv ¼ DGPB þ DGnp (4)

DEMM is the system change in molecular mechanical energyupon binding, which consists of coulombic (DECoul) and van derWaals (DEvdW) contributions, respectively. The solvation energyterm DGsolv consists of two components: the electrostatic termDGPB and the nonpolar term DGnp, respectively. DGPB is obtained bysolving the PoissoneBoltzmann equation [58] while DGnp can beobtained via the semi-empirical expression [59]:DGnp ¼ g � SASA þ b, in which SASA is the solvent accessiblesurface area of the molecule, g is the surface tension parameter(0.00542 kcal/Å2/mol), and b ¼ 0.92 kcal/mol. Finally, the entropiccontribution eTDSb is calculated via normal mode of harmonicfrequencies [60] obtained from a subset of minimized snapshotstaken from the corresponding MD trajectories.

The analysis of the energy of interaction between HSA and thedendrimers was accomplished with the MMPBSA.py script imple-mented in AmberTools14. Energy values were averaged over 200frames taken during equally spaced time interval during the last15 ns of the MD production steps. Normal mode analysis was car-ried out on a subset of 15 minimized MD snapshots evenlyextracted from the relevant trajectory time frame used for energycalculations.

E. Laurini et al. / Fluid Phase Equilibria 422 (2016) 18e3122

Finally, the effective type number of dendrimer residues andprotein amino acids involved in binding, and the correspondingcontribution to the binding free energywere obtained performing aper residue binding free energy decomposition exploiting the MDtrajectories of each given HSA/dendrimer ensemble [61]. Thisanalysis was carried out using the MM/GBSA approach [62], andwas based on the same snapshots used in the binding free energycalculations.

3. Results and discussion

3.1. Determination of viologen-phosphorous dendrimer binding siteon HSA

From structural standpoint, the countless solved x-ray crystalstructures of HSA (per se or in complex with a myriad of metabo-lites/drugs) revealed that the protein, a 585 amino acid residuemonomer, contains three homologous a-helical domains, named I,II, and III, respectively (see Fig. 2). The domains each contain tenhelices, divided into six-helix and four-helix subdomains (A and B)(see Fig. 2); the first four helices of A and B form similar anti-

Fig. 2. (Top left panel) Equilibrate molecular dynamics structure of HSA, in ribbon represenblue; IB [108e196], dark slate blue; IIA [197e297], light green; IIB [298e383], sea green; IIIAspheres; some Cl� and Naþ ions and counterions are shown as big light gray spheres and sthree different protein regions selected for the viologen-phosphorous binding site determinblue); region 3 (residues 404e582; lime green). (Right panel) Steered molecular dynamicsprotein regions 1, 2, 3, 1e2 and 1e3 (see text for details). The dendrimer is portrayed as redpanel). Water molecules, ions and counterions are omitted for clarity. (For interpretation ofthis article.)

parallel a-helix bundles.To determine the most probable binding site for the viologen-

phosphorous dendrimers 1e6 on the human serum albumin, weapplied a validated procedure based on Steered Molecular Dy-namics (SMD) simulations [35,51e56]. To the purpose, the proteinwas divided into three major regions (domains 1e3, Fig. 2, leftpanel); then, docking poses for all dendrimers where searched bySMD over these protein regions 1, 2, and 3. Furthermore, given thedimensions of some dendrimeric molecules, two further proteinzones, encompassing regions 1e2 and 2e3, respectively, wherefurther considering for SMD-guided docking operations (Fig. 2,right panel).

Calculations of the free energy of bindingDGbind for each of the 5resulting binding poses at the end of the corresponding SMDsimulation (last time frame of Fig. 2, right panel) revealed aremarkable preference for binding of all these dendrimers to HSA inthe 1e2 region (Table S1). Accordingly, all other possible bindingmodes were discarded and the detail thermodynamic analysis ofthe HSA/dendrimer binding was carried out using the most favor-able complex configuration for each HSA/dendrimer couple only.

tation colored according to the different protein subdomains: IA [residues 5e107], sky[384e407], tan; IIIB [498e582], sandy brown. Water molecules are portrayed as whitemall dim gray spheres, respectively. (Bottom left panel) Structure of HSA showing theation via SMD: region 1 (residues 5e210, dark cyan); region 2 (residues 211e403, navy(SMD) simulations to mimic the binding process of dendrimer 5 to the HSA onto the 5spheres while the HSA is in ribbon representation (coloring scheme as in bottom left

the references to color in this figure legend, the reader is referred to the web version of

E. Laurini et al. / Fluid Phase Equilibria 422 (2016) 18e31 23

3.2. Thermodynamics of binding of viologen-phosphorousdendrimers 1e6 with HSA

The results of the application of the MM-PBSA calculations tothe binding of VPDs 1e6 to HSA are presented in Table 2. Severalconsiderations can be drawn from the analysis of this Table. First, alldendrimers are endowed with high affinity for human serum al-bumin, the predicted Kb values ranging in the interval 109e106 L/mol (Table 2, last column). However, the adopted simulation pro-tocol also clearly ranks these dendrimers with respect to theirprotein binding capacity DGb in the following order: 4 > 6 > 5 > 3 >2 > 1 (Table 2, 4th column), the best HSA binder being dendrimer 4(DGb ¼ �13.40 kcal/mol) and dendrimer 1 being the least effectiveligand (DGb ¼ �9.03 kcal/mol). From a general perspective, it ap-pears that the presence of the -PO(OEt)2 substituent as the den-drimer terminal group is more beneficial than the eCHO group toHSA binding. Indeed, irrespective of the dendrimer core, overallcharge and generation, those dendrimers featuring the phospho-nate group (i.e., 3e6) are all characterized by substantially morefavorable values of DGb (Table 2). However, somewhat contrarily toexpectations, within this latter subgroup the molecules of highergeneration 5 and 6, bearing a larger number of positive charges ontheir outer shells, are predicted to be less effective than the lowergeneration counterpart 4 in binding HSA. Finally, the nature of thedendrimer core (S]P vs. N3P3), resulting in different multiplicity ofthe dendrimer branches (3 vs. 6), seems to play a role in fosteringprotein/dendrimer binding, since, considering again the subgroupof dendrimers having the same terminal group (i.e., -PO(OEt)2),lower (i.e., less negative) DGb values can be found (Figure SI1) forthose molecules featuring the S]P core (3 and 5) with respect tothose (4 and 6) having N3P3 at their focal point (vide infra and SI formore discussion on these issues).

To gain further insight into the variegated affinity of dendrimers1e6 for HSA, each global free energy of binding value wasdecomposed into its enthalpic (DHb) and entropic (-TDSb) compo-nents, as shown in the second and third columns of Table 2, fromwhich a typical enthalpy-entropy compensation behavior is infer-red (Fig. 3, left panel). In fact, for all viologen-phosphorous den-drimers considered the enthalpic change is always favorable(DHb < 0) whereas entropy variation opposes binding (-TDSb > 0).However, the entropic penalty paid by each dendrimer upon pro-tein binding is always largely overwhelmed by the correspondingenthalpic gain (Fig. 3, right panel), thereby confirming theenthalpic-driven nature of the binding mechanism of dendrimers1e6 to human serum albumin.

Further deconvolution of DGb values into its components (Eqs.(2)e(4)) reveals the overall enthalpic contribution governing thespontaneous binding of all viologen-dendrimers 1e6 onto HSA isdominated by electrostatic interactions, as expected given thehighly positively-charged nature of the dendrimers and the highly-negatively charged protein (�15e, as calculated from the aminoacid composition). Specifically, the mean value of the total elec-trostatic energy contribution to dendrimer/protein binding

Table 2Thermodynamic parameters for the binding of dendrimers 1e6 to HSA at 298 K (seetext for details).

Dendrimer DHb (kcal/mol) -TDSb (kcal/mol) DGb (kcal/mol) Kb(L/mol)

1 �12.81 ± 0.27 þ4.32 ± 0.34 �8.49 ± 0.43 1.66 � 106

2 �13.29 ± 0.25 þ4.35 ± 0.36 �8.94 ± 0.44 3.55 � 106

3 �15.06 ± 0.19 þ5.25 ± 0.31 �9.81 ± 0.36 1.54 � 107

4 �21.33 ± 0.22 þ7.08 ± 0.35 �14.25 ± 0.41 2.76 � 1010

5 �16.07 ± 0.21 þ5.59 ± 0.32 �10.48 ± 0.38 4.47 � 107

6 �17.85 ± 0.23 þ5.69 ± 0.35 �12.16 ± 0.42 8.12 � 108

(DGele ¼ DECoul þ DGPB) is negative in all cases, and substantiallymore favorable than the corresponding value of the dispersiveforces (i.e., van der Waals and hydrophobic interaction energies,DGdisp ¼ DEvdW þ DGnp), as shown in the left panel of Fig. 4.

Interestingly, however, the behavior of DGele is not mono-tonically dependent on the dendrimer charge: indeed, as high-lighted in the right panel of Fig. 4, those dendrimers of highergeneration, which bear a higher number of positively chargedgroups (i.e., 5 and 6), are only slightly more (if not less) effective inexploiting their charges upon binding HSA than their lower charge-bearing counterparts (i.e., 3 and 4). This sort of less-is-more effect[63] is not unusual in dendrimer/biomacromolecule association[34e41], and can be rationalized in terms of the capacity of eachdendrimeric structure to organize each single branch for optimalreceptor binding, as discussed in detail in the following paragraphs.

To substantiate the effects exerted by the different dendrimercharacteristics (i.e., core, charge, generation and terminal groups)on HSA binding affinity, we next processed data collected duringequilibrated MD simulations of the single molecular species in theframework of the MM/PBSA theory. Specifically, we assessed thevalues of DGb,resdend and DGb,resHSA that is, the contribution todendrimer/protein binding free energy yielded by each dendrimerbranch and protein residues, respectively (Tables SI2-5). To this aim,each dendrimer molecule has been subdivided into chemicallyconsistent fragments, as shown in the left panel of Fig. 5. The rightpanel of Fig. 5 shows equilibrated molecular dynamics snapshots ofthe corresponding dendrimer/HSA complexes in which the proteinresidues involved in binding are highlighted.

The results of this complex analysis are shown in Fig. 6 AeD.From Fig. 6 A and B (and Tables SI2-3) it can be seen that, inde-pendently of the dendrimer core, the most favorable contributionsto HSA binding stem from the external double-positive viologenfragments (BPF) and from the dendrimer terminal groups (BZA/DEP). In fact, for each dendrimer, the sum of the BPF and BZA/DEPcontributions - corresponding to �3.94, �5.15, and �4.62 kcal/molfor the S]P core-based dendrimers 1, 3, and 5, and to �4.488,�7.066, and�5.128 kcal/mol for the N3P3 core-based dendrimers 2,4, and 6, respectively (Tables SI2-3)e amount, on average, to 85% ofthe total binding free energy afforded by all dendrimer fragments(SDGb,resdend terms in Tables SI2-3). Moreover, the presence of the-PO(OEt)2 moiety (DEP) on the outer dendrimer shell is morebeneficial to protein binding than the eCHO group (BZA). In fact,the synergistic combination of -PO(Et)2 structural flexibility,essential in maximizing dendrimer/protein contacts, and higherelectronegativity, which potentiates the effect of the positivecharges on the linked BPF fragments (Fig. 6AeB), renders the BPF-DEP sequence more effective than the BPF-BZA one. Focusing theattention of those molecules featuring the -PO(OEt)2 terminalgroup (3e6), Fig. 6AeB also reveal that increasing the number ofcharged viologen branches in the dendrimeric structure does notresult in a corresponding increasing dendrimer affinity for HSA. Infact, by comparing dendrimers 3 and 5 (or 4 and 6), we see that thehigher level of dendrimer structural complexity ultimately resultsin a negligible increase (3 vs. 5) if not in a decrease (4 vs. 6) of thedendrimer ability and efficacy in protein binding (i.e., less is more).

Panels C and D in Fig. 6 show the same analysis performed onthe negatively charged HSA residues mainly involved in bindingdendrimers 1e6 (see also Tables SI4 and SI5). As it can be evincedfrom this Figure, the presence of a larger core (N3P3), with a mul-tiplicity Nc of 6, leads to dendrimers which exchange a highernumber of contacts with the protein then in the case of dendrimersfeaturing the S]P core (Nc ¼ 3). This, in turn, results in a moreoverall favorable contribution to binding. In fact, the total contri-bution afforded by the HSA glutamic and aspartic acid residues inthe case of dendrimers 1, 3 and 5 (S]P core) is equal to �2.509,

Fig. 3. (Left) Decomposition of the free energy of binding DGb of dendrimers 1e6 on HSA into its enthalpic (DHb) and entropic (-TDSb) contributions at 298 K. (Right) eTDSb vs. DHb

scatter plot for the binding of dendrimers 1e6 to HSA.

Fig. 4. (Left) Decomposition of the binding enthalpy DHb of dendrimers 1e6 in complex with HSA into overall contributions from electrostatic (DGele ¼ DECoul þ DGPB) anddispersive (DGdisp ¼ DEvdW þ DGnp) terms (Eq. (2))-4)). (Right) Overall electrostatic contribution DGele as a function of the total positive charge carried by dendrimers 1e6. Filled andopen symbols are used to distinguish between S]P core- and N3P3 core-based molecules. Data standard deviations are in the range ±0.05 ÷ ±0.15.

Fig. 5. (Left) Subdivision of VPD molecules (Fig. 1) into chemically consistent fragments (SPC: S]P core; NPC: N3P3 core; MBH: 1-methyl-2-(4-methylbenzylidene)hydrazinefragment; BPC: inner 1,10-diethyl-4,40-bipyridine-1,10-diium fragment; CCC: benzene fragment; BPF: external 1,10-diethyl-4,40-bipyridine-1,10-diium fragment (only for dendrimers 5and 6), BZA: benzaldehyde terminal group (only for dendrimers 1 and 2); DEP: -PO(OEt)2 terminal group (only for dendrimers 3e6)). (Right) Equilibrated molecular dynamicssnapshots of dendrimers 1e6 in complex with HSA. Each protein residue involved in binding is labeled. The protein is in colored ribbon representation while the dendrimermolecules are shown as colored sticks-and-balls (1, golden rod; 2, orange red; 3, yellow; 4, firebrick; 5, gold; 6 red). Hydrogen atoms, water molecules, ions and counterions are notshown for clarity. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

E. Laurini et al. / Fluid Phase Equilibria 422 (2016) 18e3124

Fig. 6. Total per residue decomposition (SDGb,resdend) of the free energy of binding DGb of dendrimer fragments for dendrimers with S]P (A) and N3P3 core (B). Per residuedecomposition (DGb,resHSA) of the free energy of binding DGb of the HSA residues mainly involved in binding dendrimers with S]P (C) and N3P3 core (D).

Table 3Molecular dynamics-based prediction of the secondary structure composition ofHSA alone and in complex with viologen-phosphorous dendrimers 1e6.

HSA alone % a-helix % Antiparallel % Parallel % Turn % Random coil

57.4 4.4 4.8 13.6 19.8

Dendrimer % a-helix % antiparallel % parallel % turn % random coil

1 55.2 4.8 5.2 14.1 20.72 55.7 5.0 4.9 14.4 20.03 57.2 4.4 4.8 13.8 19.84 57.5 4.4 4.7 13.6 19.85 56.9 4.3 4.6 14.9 20.16 57.9 4.1 4.9 13.4 19.7

E. Laurini et al. / Fluid Phase Equilibria 422 (2016) 18e31 25

�2.916, and �3.984 kcal/mol, respectively, whilst the SDGb,resHSA

values for dendrimers 2, 4, and 6, characterized by the presence ofthe N3P3 core, are equal to �2.778, �5.564, and �4.972 kcal/mol,respectively (Tables SI4 and SI5). Interestingly, however, this“protein perspective” also confirms the less is moremantra. Indeed,especially for the dendrimers with Nc ¼ 6, increasing dendrimergeneration (and, hence, structural complexity) is detrimental toprotein/dendrimer adaptation and, ultimately, effective binding.

The final result of this analysis is that dendrimer 4 representsthe best compromise in term of structure, charge optimization andterminal group; as such, it is predicted to be the most adaptive andeffective HSA binder of the entire viologen-phosphorous den-drimer series considered.

3.3. Conformational changes of HSA upon binding viologen-phosphorous dendrimers 1e6

According to the in silico analysis of the binding thermody-namics of VPDs with human serum albumin reported above, alldendrimers 1e6 should be endowed with high affinity towardsHSA. However, tight ligand/protein interactions may induceconformational changes in the polypeptide structure, which, inturn, may result in partial/substantial protein unfolding and loss offunctionality/degradation. To check whether this was the case forthe present systems, we further analyzed the molecular dynamicstrajectories of each dendrimer/HSA complex and compared thecorresponding secondary structure of the dendrimer-bound pro-tein with that obtained for the free HSA in solution.

The analysis of the secondary structural motifs of the HSA aloneand in complex with all VPDs 1e6 is reported in Table 3. As can beseen from this Table, dendrimer/HSA binding is accompanied by anegligible, if any, protein conformation change; this is particularlytrue for dendrimers 3e6, which all feature the -PO(OEt)2 terminalgroup. Structural changes in protein induced by ligand binding arecrucial mechanisms of action and regulation/deregulation of

protein biological activity. Accordingly, the preservation of HSAstructural integrity upon binding of viologen-phosphorous den-drimers 1e6 under conditions of low dendrimer concentration is avery interesting prediction, which further supports the exploitationof HSA as an effective endogenous nanocarrier for the VDPs.

3.4. Experimental validation

3.4.1. Fluorescence quenching of HSA by viologen-phosphorousdendrimers 1e6

Fluorescence techniques provide several advantages comparedwith other biophysical and biochemical methods for measuringprotein/ligand interactions. First, fluorescence intensity is linearlydependent on the number of fluorophores in a sample, providing abasis for quantitative measurements. Second, fluorescence mea-surements possess a very high sensitivity and thus can be per-formed on single molecules, providing the opportunity to observebiological mechanisms on a molecular level. Third, fluorescence is aprocess characterized by a range of different parameters, which canbe measured independently or in combination providing infor-mation not only on the mere presence of a fluorophore but also on

Table 4SterneVolmer constants for tryptophan fluorescence quenching (KSV), experimentalbinding constant (Kb), number of bound dendrimer molecules n, andexperimentally-derived free energy of binding (DGb) for viologen-phosphorousdendrimers 1e6 and HSA at 298 K. Kb and n were obtained from fluorescence-quenching data fitting using Eq. (6) while DGb was obtained from Kb using therelationship: DGb ¼ -RT ln Kb.

Dendrimer KSV (L/mol) Kb (L/mol) n DGb (kcal/mol)

1 1.84 � 105 3.49 � 106 1.057 �8.932 0.86 � 105 9.30 � 106 1.198 �9.513 3.73 � 105 2.02 � 107 1.203 �9.964 7.34 � 105 8.85 � 109 1.350 �13.595 4.08 � 105 7.21 � 107 1.113 �10.766 12.1 � 105 2.45 � 108 1.289 �11.45

E. Laurini et al. / Fluid Phase Equilibria 422 (2016) 18e3126

its orientation and its immediate environment.HSA possesses only one single tryptophan residue (W214),

located in the subdomain IIA. Accordingly, the fluorescence in-tensity behavior of HSAW214 in the presence of the different VPDs1e6 is a suitable technique to study the underlying protein/den-drimer intermolecular interactions. As expected, the progressiveaddition of each dendrimer to the protein solution resulted in theregular decrease of W214 fluorescence intensity (as exemplified fordendrimer 4 in Figure SI2), thereby confirming that all theseviologen-based dendrimers were able to quench HSA.

Fluorescence quenching is usually classified as dynamic (DQ) orstatic (SQ) quenching. Generally, the two mechanisms can bedistinguished by their different dependence on temperature.Higher temperatures result in faster diffusion and larger amountsof collision quenching, and will typically lead to the dissociation ofweakly bound complexes and smaller amount of static quenching[64]. Accordingly, the value of the quenching constant kq shouldincrease for DQ while it should decrease for SQ with increasingtemperature [65]. To determine the HSA quenching mechanismexerted by dendrimers 1e6, fluorescence-quenching data wereanalyzed using the well-known Stern-Volmer equation:

F0=F ¼ 1þ kqt0½Q � ¼ 1þ KSV½Q � (5)

in which F and F0 are the fluorescence peak intensities of the flu-orophore (i.e., W214 in HSA) in the absence and presence ofquencher, respectively, kq is the quenching rate constant of HSA, t0is the average life time of the protein (5.6 ns for HSA [66]), and [Q] isthe concentration of the quencher (dendrimers in the present case).The product of kq and t0 is also known as the SterneVolmerquenching constant KSV. The SterneVolmer plots for the quenchingof HSA by dendrimers 1e6 at room temperature are shown in Fig. 7,together with the relevant values kq, while calculated KSV are listedin the first column of Table 4.

As can be seen in Fig. 7, the values for kq calculated using Eq. (5)are in the range 1012e1013 M�1 s�1. Since all kq values are muchlarger than the maximum collisional quenching constant(2.06 � 1010 M�1 s�1), SQ is the dominant quenching mechanismfor all dendrimer/HSA complexes [67]. To further verify this asser-tion, the dependence of kq on temperature was investigated.Accordingly, the SterneVolmer plots for the quenching of HSA bydendrimers 1e6 at two further temperatures (i.e., 303, and 310 K)

Fig. 7. SterneVolmer plots and kq values for the quenching of HSA by dendrimers 1e6 at 2

were determined (Table SI6). The inverse correlation of kq withtemperature supports the evidence that the fluorescence quench-ing of HSA by all was initiated by complex formation between HSAdendrimers 1e6 rather than by dynamic collision between the twomacromolecules [68,69]. The static nature of HSA fluorescencequenching by dendrimers 1e6 was obtained from time-resolvedfluorescence analysis. Fig. 8 shows the fluorescence decay of HSAin the absence and presence of the viologen-phosphorous

98 K. [Dendrimer]: 1, 2, 3, 5, 7, 10, 12, 14, 16, 18, and 20 � 10�6 M [HSA] ¼ 5 � 10�6 M.

Fig. 8. Time-resolved fluorescence decays of HSA (1 � 10�6 mol/L) alone and in thepresence of the viologen-phosphorous dendrimer 6 (5 � 10�6 M).

Table 5Binding parameters of dendrimers 1e6 and HSA at different temperatures. Thevalues of Kb and n were obtained from fluorescence quenching data fitting using Eq.(6). DGb is obtained from Kb from the relationship: DGb ¼ -RT ln Kb.

Dendrimer T (K) Kb (L/mol) n R2 DGb (kcal/mol)

1 303 2.10 � 106 1.029 0.987 �8.63310 1.44 � 106 1.032 0.990 �8.41

2 303 5.50 � 106 1.268 0.989 �9.20310 3.72 � 106 1.163 0.994 �8.98

3 303 1.14 � 107 1.234 0.993 �9.63310 7.46 � 106 1.199 0.989 �9.38

4 303 4.21 � 109 1.330 0.988 �13.14310 2.43 � 109 1.361 0.986 �12.81

5 303 3.96 � 107 1.098 0.990 �10.37310 2.54 � 107 1.078 0.991 �10.11

6 303 1.32 � 108 1.323 0.989 �11.08310 8.36 � 107 1.304 0.990 �10.81

E. Laurini et al. / Fluid Phase Equilibria 422 (2016) 18e31 27

dendrimer 6 as an example.As it can be seen, HSA exhibits a single exponential decay in both

cases; moreover, by increasing dendrimer concentration practicallyno change in HSA lifetime was observed (i.e., from 5.54 ns to5.46 ns). Utterly analogous profiles and behaviors were observedfor all other VPDs considered (data not shown).

For static quenching and under the assumption that there are nequivalent and independent binding sites for the quencher (Q) onthe protein, the binding constant Kb for all viologen-phosphorousdendrimers 1e6 onto HSA can be calculated according to themodified SterneVolmer equation:

log½ðF0 � FÞ=F� ¼ log Kb þ n log ½Q � (6)

where F and F0 denote steady-state fluorescence of HSA with andwithout quencher, respectively, Kb is the binding constant, n is thenumber of quencher molecules bound to HSA, and [Q] is the con-centration of the quencher (again dendrimers in the present case).The linear plots of log[(F0eF)/F] vs. log [Q] obtained for all den-drimers 1e6 at 298 K are shown in Fig. 9, while the correspondingvalues of Kb and n are listed in the third and fourth columns ofTable 4.

The same analysis was repeated for data obtained at 303 K and310 K (Figure SI3), and the corresponding Kb and n values for alldendrimers considered are gathered in Table 5.

It can be seen from Tables 4 and 5 that supramolecular com-plexes between HSA and dendrimers 1e6 can be easily formed at alltemperatures; also, the number of binding sites n is always close to1, indicating that there is only 1 preferential dendrimer binding siteon the protein surface. Moreover, the Kb values presented inTables 4 and 5 reveal that the fluorescence-derived experimentalaffinity of the present dendrimer family towards HSA follows theorder: 4 > 6 > 5 > 3 > 2 > 1, in full agreement with computationalpredictions (see Table 2). In other words, dendrimer 4 is thestrongest HSA binder (DGb ¼ �13.59, �13.14, and �12.81 kcal/mol)whereas dendrimer 1 is endowed with the lowest protein bindingcapacity (DGb ¼ �8.93, �8.63, and �8.41 kcal/mol) in the temper-ature range of physiological interest considered.

In the limited range of temperature values such as thoseconsidered in this work (i.e., DT ¼ 12 K), the variation of enthalpywith temperature can be neglected, so that the enthalpy value forthe binding of dendrimers 1e6 to HSA can be calculated by applying

Fig. 9. Modified Stern-Volmer plot for the quenching of HSA by viologen-pho

van’ t Hoff equation:

ln

Kb;2

Kb;1

!¼ DHb

R

�1T1

� 1T2

�(7)

in which R is the gas constant (1.987 cal/(mol K), T1 and T2 are twoabsolute temperature values (K), and Kb,1 and Kb,2 are the bindingconstant at T1 and T2, respectively.

Thus, choosing T1 ¼ 298 K and T2 ¼ 310 K, and inserting thecorresponding values of Kb,1 and Kb,2 (Table 5) into Eq. (7), therelevant value of DHb can be estimated. Also, using the corre-sponding DGb values in Table 4, the entropic component TDSb canbe readily obtained from Eq. (8) as:

TDSb ¼ DHb � DGb (8)

Table 6 shows the results of this analysis. As it can be seen, theexperimentally-derived enthalpic and the entropic component ofthe free energy of binding DGb of dendrimers 1e6 onto HSA arenegative at all temperatures, implying that the binding processbetween all dendrimers and the protein is predominantlyenthalpic, while the entropic variation opposes binding, as mostoften seen for ligand/protein complexation and discussed in therelevant modeling section. Also, the opposite sign of DHb and TDSbconfirm that electrostatic intermolecular interactions are

sphorous dendrimers 1e6 (panels A to F) at 25 �C. [HSA] ¼ 5 � 10�6 M.

Table 6Enthalpy (DHb), entropy (TDSb), and free energy of binding (DGb) for dendrimers1e6 in complex with HSA at different temperatures. DHb and TDSb values wereobtained via Eqs. (7) and (8), using the Kb and DGb values reported in Table 5 (seetext for details).

Dendrimer T (K) DHb (kcal/mol) -TDSb (kcal/mol) DGb (kcal/mol)

1 298 �13.52 þ4.59 �8.93303 þ4.89 �8.63310 þ5.11 �8.41

2 298 �14.02 þ4.51 �9.51303 þ4.82 �9.20310 þ5.04 �8.98

3 298 �15.24 þ5.28 �9.96303 þ5.61 �9.63310 þ5.86 �9.38

4 298 �19.79 þ6.20 �13.59303 þ6.65 �13.14310 þ6.98 �12.81

5 298 �15.98 þ5.22 �10.76303 þ5.61 �10.37310 þ5.87 �10.11

6 298 �16.45 þ5.00 �11.45303 þ5.37 �11.08310 þ5.64 �10.81

E. Laurini et al. / Fluid Phase Equilibria 422 (2016) 18e3128

prevalently leading the HSA/dendrimer binding process - asanticipated by modeling results and expected given the nature ofthe receptor and its ligands.

3.4.2. Thermodynamics of binding of HSA and viologen-phosphorous dendrimers 1e6 via isothermal titration calorimetry(ITC)

ITC is a highly sensitive technique, whichmeasures the enthalpychange of a reacting system as a function of reactant amount. It isrecognized as one of the most popular techniques used for inves-tigating intermolecular and/or intramolecular interactions duringthe past decade [70]. It simultaneously gives the enthalpy change(DHb), entropy change (TDSb), and Gibbs energy change (DGb) of anassociation or a disassociation process, the binding affinity (Kb) andthe number of binding sites (n) of the receptor-ligand binding re-action, and the type of interactions depending on the exothermic orendothermic phenomena observed in the ITC experiments.Accordingly, ITC experiments were performed to determine thebinding thermodynamics of viologen-phosphorous dendrimers

Fig. 10. (Insert: ITC raw data for the interaction between the viologen-phosphorousdendrimer 4 and HSA at room temperature in PBS buffer at pH ¼ 7.4). Plot of the in-tegrated area under each ITC peak for the same system. The solid line is model datafitting with adjustable parameters n, Kb, and DHb (see text for details).

1e6 to human serum albumin at room temperature. A represen-tative ITC curve is shown in Fig. 10 for the binding of dendrimer 6with HSA, whereas the full set of ITC results is listed in Table 7.

In agreement with simulation predictions and fluorescencespectroscopy evidences, ITC data yield values of the binding con-stant Kb for all VPDs onto HSA in the range 106e109 L/mol, soconfirming that the binding affinity of all dendrimers for this pro-tein is very high. As verified by ITC, the values of the parameter nare also close to 1 for all VPD/HSA complexes (Table 7, secondcolumn), thereby validating the underlying computationalassumption of a 1:1 stoichiometry for each protein/dendrimer as-sembly formation. Most importantly, however, the ITC estimatedvalues of the binding free energyDGb and its enthalpic and entropiccontributions (DHb and eTDSb, respectively) not only show thesame qualitative trend but are also in excellent quantitativeagreement with those derived from the application of the MM/PBSA computational ansatz, as can be inferred from a comparison ofthe corresponding values listed in Tables 4 and 7 The first threepanels of Figure SI4 give a graphical portrait of this nice correlation.This allows us to conclude that the MM-PBSA method is reliablyreproducing the free energy of binding of the six viologen-phosphorous dendrimers 1e6 with human serum albumin andperforms remarkably well for rank-ordering the differential affinityof each VPD towards this protein.

3.4.3. Circular dichroism (CD) spectroscopy analysisCircular dichroism (CD) spectroscopy is awidely usedmethod to

determine protein conformation, structure and stability in solution,and to monitor interactions between proteins and other molecules.Accordingly, CD spectra of HSA alone and in the presence ofincreasing concentrations of dendrimers 1e6 were recorded todecipher the structural and conformational changes, if any, exertedby dendrimer binding on the secondary structure of the protein.Fig. 11 shows the results obtained for the system HSA/4 as a proof-of-concept. As seen in this Figure, the CD spectrum of native al-bumin exhibits two negative absorption bands in the far-UV region,peaking at 208 nm (p-p*) and 222 nm (n-p*), characteristic of thea-helix portions of the protein [71]. The addition of dendrimer 4 tothe HSA solution results in minimal reduction in band intensity atall wavelengths of the relevant CD spectra, without any discernibleshift of the band maxima. This increase in ellipticity suggests astabilization of the complex with respect to free HSA. In otherwords, the intermolecular interactions between the dendrimersand HSA do not substantially perturb the a-helical motifs of theprotein to an extent to leave a signature on the relevant CD spectra.This, in turn, implies that the dendrimers molecules cannot pene-trate inside the structure of HSA but, rather, they bind to the proteinsurface, in agreement with the molecular modeling predictionsdiscussed in paragraphs 3.2 and 3.3. This somewhat expected resultcould be further rationalized considering that HSA is a singlepolypeptide chain characterized by the presence of 17 disulfidebridges, located at regular intervals along the protein structure.Accordingly, in addition to its general compact folded nature, HSAflexibility is considerably restricted by virtue of this network of SeSlinks. Hence, when dendrimer 4 interacts with HSA in solution, theprotein native structure is not expected to unfold to a considerableextent due to 4 surface binding.

Table 8 shows the analysis of the secondary structural motifs ofthe HSA per se and in complex with VPDs 1e6 as extracted from thecorresponding CD data obtained at the lowest dendrimer/HSAmolar ratio (see Table SI7 for data at all molar ratios). From thistable it is clear that all six dendrimers do not induce any significantconformational change in the protein for all values of the den-drimer/HSAmolar ratio (MR). Pleasingly, this result agrees with theHSA secondary structure analysis predicted by computer

Table 7Thermodynamic parameters for the binding of dendrimers 1e6 to HSA at 298 K (see text for details).

Dendrimer n Kb(L/mol) DHb (kcal/mol) -TDSb (kcal/mol) DGb (kcal/mol)

1 1.104 4.15 � 106 �13.26 þ4.23 �9.032 1.273 6.26 � 106 �13.65 þ4.38 �9.273 1.185 1.35 � 107 �14.79 þ5.06 �9.734 1.298 6.53 � 109 �20.27 þ6.87 �13.405 1.109 1.01 � 108 �16.48 þ5.56 �10.926 1.071 5.27 � 108 �17.04 þ5.14 �11.90

Fig. 11. Example of far UV-CD spectra of HSA in the presence of increasing concen-trations of VPD 4 at room temperature. MRE ¼ mean residue ellipticity. 4/HSA molarratio from a to e: 0, 1, 5, 10, and 20.

Table 8Composition of the secondary structure of HSA alone and in complex with VPDs1e6 at dendrimer/HSA molar ratio ¼ 1 as determined from CD measurements.

HSA alone % a-helix % Antiparallel % Parallel % Turn % Random coil

58.0 4.1 4.5 13.4 20.0

Dendrimer % a-helix % antiparallel % parallel % turn % random coil

1 54.7 4.9 5.5 14.3 20.62 55.0 4.8 5.2 14.6 20.43 57.8 4.2 4.3 14.1 19.64 57.3 4.1 4.9 14.0 19.75 56.2 4.8 4.2 13.9 20.96 57.4 4.3 4.5 13.8 20.0

E. Laurini et al. / Fluid Phase Equilibria 422 (2016) 18e31 29

simulations, thereby constituting a further cornerstone of valida-tion of the modeling results.

A closer inspection of Table SI7 however reveals that, startingfrom MR ¼ 5, some structural perturbations indeed become man-ifest, which reflect into a decrease in the a-helical content and acorresponding increase of the b-turn/random coil motifs. Notably,these effects are observed only for the smallest dendrimers 1 and 2,while a negligible influence e if any - on the HSA secondarystructure is observed for dendrimers 3e6. A likely explanation forthis evidence may be that, in the case dendrimers 1 and 2, a den-drimer concentration of 5 mM is high enough to foster furtherdendrimer/protein unspecific binding, which, in turn, induces aslightly higher degree of conformational perturbation in the humanserum albumin structure.

4. Conclusions

Dendrimers constitute a class of perfectly defined, self-similar

macromolecules, whose architecture consists in a central core,from which identical branches emanate radially, and compriseidentical layers, termed generations. One of the most interestingaspects of dendrimers resides in the so-called dendrimer (or den-dritic) effect [9], which is observed when a specific chemical func-tion exhibit different properties, depending onwhether it is graftedor not to a dendrimer scaffold. Specifically, if all generations of agiven dendrimer are endowed with the same properties but themonomer does behave differently, then this effect is defined asmultivalency. On the other hand, if the considered property varieswith dendrimer generation, the phenomenon is called generationeffect, the underlying consequence being most often ascribable tothe dendrimer terminal groups.

Dendrimers featuring 4,40-bipyridinium salts (i.e., viologens) assubunits have recently become very attractive due to the peculiarproperties of this chemical moiety. In fact, viologens are well-known electrochromic materials, and can form strong chargee-transfer complexes with a variety of suitable donor compounds[72]. Thus, a plethora of dendrimers bearing the viologen group atthe core, along the branches, or at their periphery have been syn-thesized, characterized, and substantial efforts have been devotedto the optimization of the relevant synthetic strategies, specificfunctionalization, and physical and biological applications [73e75].

Phosphorous is an element that plays a key role in almost allaspects of life. Thus, in the form of the triple-negatively chargephosphate anion it is a fundamental constituent of nucleic acids,phosphorylation is themain starting event in the cast majority of allcellular signaling cascades, mixtures of phospholipids constitutecellular membranes and, last but not least, inorganic phosphatesare vital components of bones. It should be not surprising then thatchemicals containing phosphorous in their formula may interfere(either positively or negatively) with biological systems. Accord-ingly, the idea of incorporating phosphorous into the chemicalstructure of dendrimers resulted in phosphorous-containing den-drimers, which were proven to exhibit a variety of roles when incontact with biomolecular entities [13].

Human serum albumin constitutes some 50% of the proteinpresent in the plasma of normal healthy individuals. Albumin is themain determinant of plasma oncotic pressure and it plays a pivotalrole in modulating the distribution of fluids between compart-ments. Moreover, it has many other biological properties whichmay be important not only for its physiological actions but also forits therapeutic effects. Among these are its capacity of moleculetransportation and free radical scavenging, its ability to modulatecapillar permeability, neutrophil adhesion and activation, and itshemostatic effects [1].

Notwithstanding the potential perspective of biomedical ap-plications of viologen-phosphorus dendrimers outlined above,unexpectedly no studies beyond basic biological properties of VPDshave been published so far. Therefore, with the aim of exploitingHSA as a potential carrier for VPDs, in this work we performed apreliminary investigation of the interaction of the six differentVPDs 1e6 with HSA using a combination of computational andexperimental techniques. Accordingly, different simulations

E. Laurini et al. / Fluid Phase Equilibria 422 (2016) 18e3130

methodologies were employed to determine the dendrimer/HSAbinding mode and the relevant binding thermodynamic parame-ters. All molecular determinants contributing to VPDs/HSA freeenergy of binding DGb were analyzed, and the eventual HSAstructural variations induced by dendrimer binding were quanti-fied. Finally, all modeling predictions were validated using a seriesof experimental approaches, including isothermal titration calo-rimetry (ITC), circular dichroism (CD), and fluorescence quenchingand decay (see also last Panel in Figure SI4). In summary, the resultsfrom this study allowed us to rank the affinity of the differentviologen-phosphorous dendrimers 1e6 towards HSA and toformulate a molecular-based rationale for the differential bindingthermodynamics of the resulting dendrimer/HSA complexes. Ac-cording to our data, HSA can successfully and selectively bind VPDs1e6, dendrimer 4 being the best cargo for this endogenous proteinnanocarrier.

Acknowledgments

The financial support from the University of Trieste, FRA project“Combined computational/experimental evidences for multivalent,amphiphilic nanoscale carriers” is gratefully acknowledged.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.fluid.2016.02.014.

References

[1] T. Peter Jr. (Ed.), All about Albumin: Biochemistry, Genetics and Medical Ap-plications, Academic Press, San Diego and London, 1996.

[2] M. Fasano, S. Curry, E. Terreno, M. Galliano, G. Fanali, P. Narciso, S. Notari,P. Ascenzi, The extraordinary ligand binding properties of human serum al-bumin, IUMBM Life 57 (2005) 787e795.

[3] F. Kratz, Albumin as drug carrier: design of prodrugs, drug conjugates andnanoparticles, J. Control. Release 132 (2008) 171e183.

[4] P. Ascenzi, M. Fasano, Serum heme-albumin: an allosteric protein, IUBMB Life61 (2009) 1118e1122.

[5] G. Fanali, A. di Masi, V. Trezza, M. Marino, M. Fasano, P. Ascenzi, Human serumalbumin: from bench to bedside, Mol. Asp. Med. 33 (2012) 209e290.

[6] D.A. Tomalia, Birth of a new macromolecular architecture: dendrimers asquantized building blocks for nanoscale synthetic polymer chemistry, Prog.Polym. Sci. 30 (2005) 294e324.

[7] C.L. Cameron, J.A. MacKay, J.M.J. Fr�echet, F.C. Szoka, Designing dendrimers forbiological applications, Nat. Biotechnol. 23 (2005) 1517e1526.

[8] D.A. Tomalia, Dendrons/dendrimers: quantized, nano-element like buildingblocks for soft-soft and sof-hard nano-compound synthesis, Soft Matter 6(2010) 456e474.

[9] D.A. Tomalia, Dendritic effect: dependency of dendritic nano-periodic prop-erty patterns on critical nanoscale design parameters (CNDPs), New J. Chem.36 (2012) 264e281.

[10] A.M. Caminade, C.O. Turrin, R. Laurent, A. Ouali, B. Delavaux-Nicot (Eds.),Dendrimers. Towards Catalytic, Material and Biomedical Uses, Wiley, NewYork, 2011.

[11] N. Launay, A.M. Caminade, R. Lahana, A general strategy for neutralphosphorous-containing dendrimers, Angew. Chem. Int. Engl. Ed. 33 (1994)1589e1592.

[12] A.M. Caminade, A. Ouali, R. Laurent, C.O. Turrin, J.P. Majoral, The dendriticeffect illustrated with phosphorous dendrimers, Chem. Soc. Rev. 44 (2015)3890e3899.

[13] A.M. Caminade, C.O. Turrin, J.P. Majoral, Biological properties of phosphorousdendrimers, New J. Chem. 34 (2010) 1512e1524.

[14] C. Huang, H. Zhang, Y. Jiang, G. Li, H. Wang, X. Tang, Q. Wang, Paraquatinduced convulsion and death: a report of five cases, Toxicol. Ind. Health 29(2013) 722e727.

[15] T. Fukushima, K. Tanaka, H. Lim, M. Moriyama, Mechanism of cytotoxicity ofParaquat, Environ. Health Prev. Med. 7 (2002) 89e94.

[16] C. Freire, S. Koifman, Pesticide exposure and Parkinson's disease: epidemio-logical evidence of association, Neurotoxicology 33 (2012) 947e971.

[17] C. Padi�e, M. Maszewska, K. Majchrzak, B. Nawrot, A.M. Caminade, J.P. Majoral,Polycationic phosphorous dendrimers: synthesis, characterization, study oftoxicity, complexation of DNA, and transfection experiments, New J. Chem. 33(2009) 318e326.

[18] A.M. Caminade, J.P. Majoral, Positively charged phosphorous dendrimers. an

overview of their properties, New J. Chem. 37 (2013) 3358e3373.[19] K. Murugavel, Benzylic viologen dendrimers: a review of their synthesis,

properties and applications, Polym. Chem. 5 (2014) 5873e5884.[20] S. Asaftei, E. De Clercq, “Viologen” dendrimers as antiviral agents: the effect of

charhe number and distance, J. Med. Chem. 53 (2010) 3480e3488.[21] K. Ciepluch, N. Katir, A. El Kadib, A. Felczak, K. Zawadzka, M. Weber,

B. Klajnert, K. Lisowska, A.-M. Caminade, M. Bousmina, M. Bryszewska,J.P. Majoral, Biological properties of new viologen-phosphorous dendrimers,Mol. Pharm. 9 (2012) 448e457.

[22] N. Katir, J.P. Majoral, A. El Kadib, A.M. Caminade, M. Bousmina, Molecular andmacromolecular engineering with viologens as building blocks: rationaldesign of phosphoruseviologen dendritic structures, Eur. J. Org. Chem. 2(2012) 269e273.

[23] K. Ciepluch, M. Weber, N. Katir, A.M. Caminade, A. El Kadib, B. Klajnert,J.P. Majoral, M. Bryszewska, Effect of viologenephosphorus dendrimers onacetylcholinesterase and butyrylcholinesterase activities, Int. J. Biol. Macro-mol. 54 (2013) 119e124.

[24] K. Milowska, J. Grochowina, N. Katir, A. El Kadib, J.P. Majoral, M. Bryszewska,T. Gabryelak, Viologen-phosphorous dendrimers inhibit a-synuclein fibrilla-tion, Mol. Pharm. 10 (2013) 1131e1137.

[25] C. Galliot, A.-M. Caminade, F. Dahan, J.P. Majoral, Synthesis, structure, andreactivity of stable PN heterocycles with two and six methyleneamine units:[H2CNN(Me)2P(S)(Ph)] and (H2CNN(Me)]6P3N3, Angew. Chem. Int. Ed. Engl. 32(1993) 1477e1479.

[26] S. Jana, S. Ghosh, S. Dalapati, N. Guchhait, Exploring structural changes ofprotein bovine serum albumin by external perturbation using extrinsic fluo-rescence probe: spectroscopic measurement, molecular docking and molec-ular dynamics simulation, Photochem. Photobiol. Sci. 11 (2012) 323e332.

[27] L. Whitmore, B.A. Wallace, Protein secondary structure analyses from circulardichroism spectroscopy: methods and reference databases, Biopolymers 5(2008) 392e400.

[28] J.G. Lees, A.J. Miles, R.W. Janes, B.A. Wallace, Novel methods for secondarystructure determination using low wavelength (VUV) circular dichroismspectroscopic data, BMC Bioinforma. 7 (2006) 507e517.

[29] I.H.M. Van Stokkum, H.J.W. Spoelder, M. Bloemendal, R. Van Grondelle,F.C.A. Groen, Estimation of protein secondary structure and error analysisfrom circular dichroism spectra, Anal. Biochem. 191 (1990) 110e118.

[30] L.A. Compton, W.C. Johnson Jr., Analysis of protein circular dichroism spectrafor secondary structure using a simple matrix multiplication, Anal. Biochem.155 (1986) 155e167.

[31] N. Sreerama, R.W. Woody, Estimation of protein secondary structure fromcircular dichroism spectra: comparison of CONTIN, SELCON, and CDSSTRmethods with an expanded reference set, Anal. Biochem. 287 (2000)252e260.

[32] D.A. Case, V. Babin, J.T. Berryman, R.M. Betz, Q. Cai, D.S. Cerutti,T.E. Cheatham III, T.A. Darden, R.E. Duke, H. Gohlke, A.W. Goetz, S. Gusarov,N. Homeyer, P. Janowski, J. Kaus, I. Kolossv�ary, A. Kovalenko, T.S. Lee,S. LeGrand, T. Luchko, R. Luo, B. Madej, K.M. Merz, F. Paesani, D.R. Roe,A. Roitberg, C. Sagui, R. Salomon-Ferrer, G. Seabra, C.L. Simmerling, W. Smith,J. Swails, R.C. Walker, J. Wang, R.M. Wolf, X. Wu, P.A. Kollman, AMBER 14,University of California, San Francisco, USA, 2014.

[33] http://www.esteco.com/modefrontier.[34] T. Wei, C. Chen, J. Liu, C. Liu, P. Posocco, X. Liu, Q. Cheng, S. Huo, Z. Liang,

M. Fermeglia, S. Pricl, X.J. Liang, P. Rocchi, L. Peng, Anticancer drug nano-micelles formed by self-assembling amphiphilic dendrimer to combat cancerdrug resistance, Proc. Natl. Acad. Sci. U. S. A. 10 (2015) 2978e2983.

[35] D. Marson, E. Laurini, P. Posocco, M. Fermeglia, S. Pricl, Cationic carbosilanedendrimers and oligonucleotide binding: an energetic affair, Nanoscale 7(2015) 3876e3887.

[36] X. Liu, J. Zhou, T. Yu, C. Chen, K. Sengupta, Y. Huang, H. Li, C. Liu, Y. Wang,P. Posocco, M. Wang, Q. Cui, S. Giorgio, M. Fermeglia, F. Qu, S. Pricl, Y. Shi,Z. Liang, P. Rocchi, J.J. Rossi, L. Peng, Adaptive amphiphilic dendrimer-basednanoassemblies as robust and versatile siRNA delivery systems, Angew.Chem. Int. Ed. Engl. 53 (2014) 11822e11827.

[37] P. Posocco, X. Liu, E. Laurini, D. Marson, C. Chen, C. Liu, M. Fermeglia, P. Rocchi,S. Pricl, L. Peng, Impact of siRNA overhangs for dendrimer-mediated siRNAdelivery and gene silencing, Mol. Pharm. 10 (2013) 3262e3273.

[38] P. Posocco, E. Laurini, V. Dal Col, D. Marson, K. Karatasos, M. Fermeglia, S. Pricl,Tell me something I do not know. Multiscale molecular modeling of den-drimer/dendron organization and self-assembly in gene therapy, Curr. Med.Chem. 19 (2012) 5062e5087.

[39] X. Liu, C. Liu, E. Laurini, P. Posocco, S. Pricl, F. Qu, P. Rocchi, L. Peng, Efficientdelivery of sticky siRNA and potent gene silencing in a prostate cancer modelusing a generation 5 triethanolamine-core PAMAM dendrimer, Mol. Pharm. 9(2012) 470e481.

[40] X. Liu, J. Wu, M. Yammine, J. Zhou, P. Posocco, S. Viel, C. Liu, F. Ziarelli,M. Fermeglia, S. Pricl, G. Victorero, C. Nguyen, P. Erbacher, J.P. Behr, L. Peng,Structurally flexible triethanolamine core PAMAM dendrimers are effectivenanovectors for DNA transfection in vitro and in vivo to the mouse thymus,Bioconjug. Chem. 22 (2011) 2461e2473.

[41] A. Barnard, P. Posocco, S. Pricl, M. Calderon, R. Haag, M.E. Hwang, V.W. Shum,D.W. Pack, D.K. Smith, Degradable self-assembling dendrons for gene de-livery: experimental and theoretical insights into the barriers to cellular up-take, J. Am. Chem. Soc. 133 (2011) 20288e20300.

[42] J. Wang, R.M. Wolf, J.W. Caldwell, P.A. Kollman, D.A. Case, Development and

E. Laurini et al. / Fluid Phase Equilibria 422 (2016) 18e31 31

testing of a general amber force field, J. Comput. Chem. 25 (2004) 1157e1174.[43] M.W. Schmidt, K.K. Baldridge, J.A. Boatz, S.T. Elbert, M.S. Gordon, J.H. Jensen,

S. Koseki, N. Matsunaga, K.A. Nguyen, S. Su, T.L. Windus, M. Dupuis,J.A. Montgomery, General atomic and molecular electronic structure system,J. Comput. Chem. 14 (1993) 1347e1363.

[44] M.S. Gordon, M.W. Schmidt, Advances in electronic structure theory: GAMESSa decade later, in: Elsevier (Ed.), Theory and Applications of ComputationalChemistry: the First Forty Years, Elsevier, Amsterdam, 2005, pp. 1167e1189.

[45] S. Sugio, A. Kashima, S. Mochizuki, M. Noda, K. Kobayashi, Crystal structure ofhuman serum albumin at 2.5 A resolution, Protein Eng. 12 (1999) 439e446.

[46] W.L. Jorgensen, J. Chandrasekhar, J.D. Madura, R.W. Impey, M.L. Klein, Com-parison of simple potential functions for simulating liquid water, J. Chem.Phys. 79 (1983) 926e935.

[47] J.-P. Ryckaert, G. Ciccotti, H.J.C. Berendsen, Numerical integration of the car-tesian equations of motion of a system with constraints: molecular dynamicsof n-alkanes, J. Comput. Phys. 23 (1977) 327e341.

[48] X. Wu, B.R. Brooks, Self-guided Langevin dynamics simulation method, Chem.Phys. Lett. 381 (2003) 512e518.

[49] H.J.C. Berendsen, J.P.M. Postma, W.F. van Gunsteren, A. DiNola, J.R. Haak,Molecular dynamics with coupling to an external bath, J. Chem. Phys. 81(1984) 3684e3690.

[50] T. Darden, D. York, L. Pedersen, Particle mesh Ewald: An N,log(N) method forEwald sums in large systems, J. Chem. Phys. 98 (1993) 10089e10092.

[51] S. Pricl, B. Cortelazzi, V. Dal Col, D. Marson, E. Laurini, M. Fermeglia, L. Licitra,S. Pilotti, P. Bossi, F. Perrone, Smoothened (SMO) receptor mutations dictateresistance to vismodegib in basal cell carcinoma, Mol. Oncol. 9 (2015)389e397.

[52] E. Laurini, P. Posocco, M. Fermeglia, D.L. Gibbons, A. Quintas-Cardama, S. Pricl,Through the open door: Preferential binding of dasatinib to the active form ofBCR-ABL unveiled by in silico experiments. Mol. Oncol. 7 (2013) 968e975.

[53] F. Bozzi, E. Conca, E. Laurini, P. Posocco, A. Lo Sardo, G. Jocoll�e, R. Sanfilippo,A. Gronchi, F. Perrone, E. Tamborini, G. Pelosi, M.A. Pierotti, R. Maestro, S. Pricl,S. Pilotti, In vitro and in silico studies of MDM2/MDMX isoforms predictNutlin-3A sensitivity in well/de-differentiated liposarcomas, Lab. Invest 93(2013) 1232e1240.

[54] S. Park, K. Schulten, Calculating potentials of mean force from steered mo-lecular dynamics simulations, J. Chem. Phys. 120 (2004) 5946e5961.

[55] S. Park, F. Khalili-Araghi, E. Tajkhorshid, K. Schulten, Free energy calculationfrom steered molecular dynamics simulations using Jarzynski's equality,J. Chem. Phys. 119 (2003) 3559e3566.

[56] B. Isralewitz, J. Baudry, J. Gullingsrud, D. Kosztin, K. Schulten, Steered mo-lecular dynamics investigations of protein function, J. Mol. Graph. Model 19(2001) 13e25.

[57] J. Srinivasan, T.E. Cheatham, P. Cieplak, P.A. Kollman, D.A. Case, Continuumsolvent studies of the stability of DNA, RNA, and phosphoramidate�DNAhelices, J. Am. Chem. Soc. 120 (1998) 9401e9409.

[58] M.K. Gilson, K.A. Sharp, B.H. Honig, Calculating the electrostatic potential of

molecules in solution e method and error assessment, J. Comput. Chem. 9(1988) 327e335.

[59] D. Sitkoff, K.A. Sharp, B.H. Honig, Accurate calculation of hydration free en-ergies using macroscopic solvent models, J. Phys. Chem. 98 (1994)1978e1988.

[60] I. Andricioaei, M. Karplus, On the calculation of entropy from covariancematrices of the atomic fluctuations, J. Chem. Phys. 115 (2001) 6289e6292.

[61] A. Onufriev, D. Bashford, D.A. Case, Modification of the Generalized Bornmodel suitable for macromolecules, J. Phys. Chem. B 104 (2000) 3712e3720.

[62] M. Feig, A. Onufriev, M.S. Lee, W. Im, D.A. Case, C.L. Brooks III, Performancecomparison of generalized born and Poisson methods in the calculation ofelectrostatic solvation energies for protein structures, J. Comput. Chem. 25(2004) 265e284.

[63] P. Posocco, S. Pricl, S.P. Jones, A. Barnard, D.K. Smith, Less is more - multiscalemodelling of self-assembling multivalency and its impact on DNA binding andgene delivery, Chem. Sci. 1 (2010) 393e404.

[64] A. Ray, B.K. Seth, U. Pal, S. Basu, Nickel (II)-Schiff base complex recognizingdomain II of bovine and human serum albumin: spectroscopic and dockingstudies, Spectrochim. Acta Part A 92 (2012) 164e174.

[65] X.C. Zhao, R.T. Liu, Z.X. Chi, Y. Teng, P.F. Qin, New insights into the behavior ofbovine serum albumin adsorbed onto carbon nanotubes: comprehensivespectroscopic studies, J. Phys. Chem. B 114 (2010) 5625e5631.

[66] N. Tayeh, T. Rungassamy, J.R. Albani, J. Pharm, Biomed. Anal. 50 (2009)107e116.

[67] G. Zhang, Q. Que, J. Pan, J. Guo, Study of the interaction between icariin andhuman serum albumin by fluorescence spectroscopy, J. Mol. Struct. 881(2008) 132e138.

[68] R. Punith, J. Seetharamappa, Spectral characterization of the binding andconformational changes of serum albumins upon interaction with an anti-cancer drug, anastrozole, Spectrochim. Acta Part A 9 (2012) 37e41.

[69] Y. Zhang, Z.D. Qi, C.H. Li, Y. Liu, Interactions of chromium (III) and chromium(IV) with bovine serum albumin studied by UV spectroscopy, circular di-chroism, and fluorimetry, Biol. Trace Elem. Res. 130 (2009) 172e184.

[70] K.C. Tam, E. Wyn-Jones, Insights on polymer surfactant complex structuresduring the binding of surfactants to polymers as measured by equilibrium andstructural techniques, Chem. Soc. Rev. 35 (2006) 693e709.

[71] B. Ahmad, S. Parveen, R.H. Khan, Effect of albumin conformation on thebinding of ciprofloxacin to human serum albumin: a novel approach directlyassigning binding site, Biomacromolecules 7 (2006) 1350e1356.

[72] K. Murugawel, Benzylic viologen dendrimers: a review of their synthesis,properties, and applications, Polym. Chem. 5 (2014) 5873e5884.

[73] A.M. Caminade, J.P. Majoral, Dendrimers and nanotubes: a fruitful association,Chem. Soc. Rev. 1 (2010) 2034e2047.

[74] S. El Kazzuoli, S. Mignani, M. Bousmina, J.P. Majoral, Dendrimer therapeutics:covalent and ionic attachments, New J. Chem. 6 (2012) 227e240.

[75] Y. Cheng, L. Zhao, T. Li, Dendrimeresurfactant interactions, Soft Matter 10(2014) 2714e2727.