Analyst 2017

Analyst
TUTORIAL REVIEW
Association models for binding of molecules to

Cite this: Analyst, 2017, 142, 2067
nanostructures†
Manuel Ahumada,a Eduardo Lissi,b Ana Maria Montagut,a
Francisco Valenzuela-Henríquez,c Natalia L. Pacioni*d and Emilio I. Alarcon *a,e
The interaction between nanoparticles and molecules plays a key role in determining the activity and per-
formance of a given nanostructure. These interactions are pivotal for a variety of applications including
drug delivery, surface manipulation for targeted therapies, and catalysis. However, to this day, gathering
precise association parameters for the interaction of the molecules with nanostructures remains elusive
Received 18th February 2017, and mostly imprecise. In this review, we present a critical discussion of the most commonly used tech-
Accepted 18th April 2017
niques and models intended for determining the association of molecules with nanoparticles. Particular
DOI: 10.1039/c7an00288b emphasis has been put on discussing the limitations and pitfalls related to determining association con-
rsc.li/analyst stants in this tutorial review.
a
Bio-nanomaterials Chemistry and Engineering Laboratory, Division of Cardiac
Surgery, University of Ottawa Heart Institute, 40 Ruskin Street, Rm H5229, Ottawa, 1. Introduction
Canada. E-mail: ealarcon@ottawaheart.ca
b
Laboratorio de Cinética y Fotoquímica, Departamento de Ciencias del Ambiente- The binding of low molecular weight molecules to biomacro-
Facultad de Química y Biología, Universidad de Santiago de Chile, molecules plays a crucial role in a wide variety of processes
Avenida Libertador Bernardo O’Higgins 3363, Santiago, Chile including cell metabolism and biodistribution of nutrients
c
Instituto de Matemática, Pontificia Universidad Católica de Valparaíso,
and drugs in living organisms. These significantly exemplify
Blanco Viel 596, Cerro Barón, Valparaíso, Chile
d
INFIQC-CONICET and Universidad Nacional de Córdoba, Departamento de the importance of non-covalent interactions, except transition
Química Orgánica-Facultad de Ciencias Químicas, Haya de la Torre y Medina states of enzyme–substrate complexes,1–3 in the delicate
Allende s/n, X5000HUA, Ciudad Universitaria, Córdoba, Argentina. balance of life. Understanding the driving forces and binding
E-mail: nataliap@fcq.unc.edu.ar mechanisms behind the protein–small molecule interactions
e
Department of Biochemistry, Microbiology, and Immunology, Faculty of Medicine,
took decades of multidisciplinary research, which ultimately
University of Ottawa, Ottawa, ON K1Y 4 W7, Canada
† Electronic supplementary information (ESI) available. See DOI: 10.1039/ converged in the advancement of precise pharmacological
c7an00288b tools that have had a tremendous impact on the development
Manuel Ahumada studied chem- Eduardo Lissi Gervaso obtained

istry at the University of his Chemistry degree from the
Santiago de Chile. In 2016, he Universidad de Buenos Aires and
received his Ph.D. in Chemistry then moved to the University of
under the supervision of Gales for a Ph.D. (1963). Dr Lissi
Dr Eduardo Lissi, with a disser- pioneered the field of chemical
tation on solute transport kinetics and photochemistry in
through lipid membranes. The the early 70s. Professor Lissi is a
same year he joined Drs Alarcon Full Professor at the Universidad
and Suuronen at the University de Santiago de Chile and has
of Ottawa Heart Institute, published more than 500 articles
Ottawa, Canada, as a post- in peer-reviewed journals,
Manuel Ahumada doctoral fellow. Dr Ahumada’s Eduardo Lissi written book chapters and edited
research interests cover the devel- books on physical chemistry.
opment, characterization, and application of new bionano-
materials for biomedical uses.
This journal is © The Royal Society of Chemistry 2017 Analyst, 2017, 142, 2067–2089 | 2067
Tutorial Review Analyst
of therapeutics (see ref. 3 for a comprehensive review on structures around the surface. Thus, for example, for the same
binding parameters for proteins and enzymes). In most cases, nanoparticle (size, shape, and surface composition), it is not
the association of the low molecular weight molecules occurs surprising to find discrepancies in association constants deter-
in the determined regions of the protein (i.e. pockets or mined using different techniques, see Table 2. For example,
binding sites), where specific interactions between amino acid Wang et al. reported the association of trypsin with gold nano-
residues and the guest take place. Steric factors prevent multi- particles11 using two different models: the Hill equation and
occupation from happening, a reason why a 1 : 1 stoichiometry the Stern–Volmer (SV) equation. Interestingly, when using the
for the binding process, in a site model, is a reasonable latest models, the calculated Ka was twice the one calculated
assumption under physiological conditions. Association confor Hill’s equation. Hou et al. reported similar trends with the
stants (Ka) for 1 : 1 molecule–protein binding site complexes association measured by SV ≈ 1/2 of the value calculated using
range from 103.5 to 1016 M−1 for catalytic antibodies4 and enzy- Scatchard’s plots.12 Discrepancies have also been reported for
matic transition states, respectively.2,3 However, despite the nanosilver, whereas for example, Roy et al. measured the
remarkable advancements in recombinant protein synthesis, association of tyrosine, tryptophan, and phenylalanine and
proteins in general fall short regarding their stability and pro- found that the association constants measured by using the
duction cost for being realistically considered as vehicles for Benesi–Hilderbrand approach were about one order of magni-
drug delivery. Furthermore, any modification done in the tude smaller than those measured using fluorescence quench-
protein sequence for improving the plasmatic lifetime, for ing.13 Gebauer et al. reported that the association constant for
example, might lead to immune response and other side human serum albumin (HSA) to nanosilver measured by circu-
effects.5 These shortcomings have motivated the search for an lar dichroism doubled those calculated from Hill’s plots.14
alternative and improved source of drug cargo molecules. A particular case where the association constants derived by
The unique versatility and flexibility of surface composition two different methods were similar (a modified SV plot and
and manipulation displayed by nanomaterials explain, in part, Scatchard) was reported by Jiang et al., who measured the
the increasing interest generated in recent years for using association of bovine serum albumin (BSA) with nanosilver.15
engineered materials as new avenues for drug delivery6–9 and BSA is amongst the most studied proteins for the associ-
diagnostics.7–9 When compared to standard carrier vehicles, ation of gold nanoparticles; some examples are found in ref.
see Table 1, nanoparticles exhibit some attractive properties 15–19, where again, disparities amongst the association con-
that make them excellent candidates for the targeted transport stants are evident. Even when considering the different sizes
of molecules, see for example DEP™ docetaxel6 (cancer of these particles, it is clear that the methodology chosen to
therapy for drug delivery in the Phase I Clinical Trial) and determine the binding seems to be the determining factor.
CellSearch7,9,10 (nanoparticle-antibody based detection of cir- Similar trends were found for another plasma protein like
culating tumor cells). HSA.20,21
However, concepts and models for quantifying the associ- In this tutorial review, we have tackled the problem of
ation of molecules, small and large, to nanoparticulate sur- measuring/quantifying the extent of association of molecules
faces are difficult to interpret due to the intrinsic complexity of with nanoparticles by selecting representative examples of
the binding mechanisms, which can, in many cases, involve binding data that have been recently reported in the literature.
the multi-step association and formation of supramolecular Our selection criteria comprised papers that include at least 4
Ana Maria Montagut obtained Francisco H. Valenzuela

her Bachelor of Chemistry degree obtained his degree in math-
from the Autonomous University ematics from the Universidad de
of Barcelona. During her studies, Santiago de Chile and went to
she visited the Georg-August the Instituto Nacional de
Universität, Göttingen, Germany Matemática Pura e Aplicada
and the University of Ottawa (IMPA), Brazil to complete a
Heart Institute, Ottawa, Canada Ph.D. in Mathematics (2009).
(Dr Alarcon’s group). She has a Dr Valenzuela’s interests cover
master’s degree in Industrial the study of dynamical systems
Chemistry from the Autonomous and ergodic theory. He is also
University of Barcelona. Later, currently studying the mathe-
Ana Maria Montagut she completed her Ph.D. in Francisco Valenzuela- matical interpretation of nano-
Organic Chemistry with a disser- Henriquez structured systems for bio-
tation on hydrophobic and microbiocidal materials under the medical sensing.
supervision of Dr Rosa María Sebastián and Prof. Adelina
Vallribera.
2068 | Analyst, 2017, 142, 2067–2089 This journal is © The Royal Society of Chemistry 2017
Analyst Tutorial Review
out of the following seven properties: ligand, nanoparticle the association of molecules with nanostructured systems.
size, nanoparticle shape, surface charge, polydispersity, associ- Thus, we have organized this review in separate sections.
ation model employed, and occupation number. Section 2 deals with the fundamental concepts of association
A list of nanomaterials that have been revised in this review, and factors that can interfere with it for the case of nano-
see Table 2, includes: gold (spheres,15–18,20–26 spheroids,12,27,28 materials. Section 3 is devoted to analyzing and critically dis-
rods,17,26 cubes,29,30 and others),19,31–34 silver (spheres,14,35 cussing the different methodologies and the corresponding
spheroids,15,36 polymorphs,37 and others13,38,39), carbon,40 binding models that have been used in the literature to
copper,41 polystyrene,35,42 aluminum oxide,43 cerium oxide,43 measure and quantify the association of nanoparticles. In
cobalt–iron,44,45 iron oxide,11,44,46,47 titanium oxide,43 tin section 4, we present an overview of guidelines on the steps/
oxide,48 zinc oxide,49,50 and zinc sulfur.51 As per the ligands models recommended for determining the association of
presented in our literature revision, listed in alphabetical molecules with nanoparticles. Finally, in section 5, an outlook
order: 1,4-dimethoxy-2,3-dibromomethylanthracene-9,10-dione and summary of the review are presented. For educational pur-
(DMDBMAD),36 3-(10-hexyloxyethyl)-3-devinylpyropheophor- poses, and to highlight differences and similarities, we have
bide-a (HPPH),29 [32]ane-(NH2+)8.8Cl−,38 α-synuclein,32 anti- used proteins as benchmarks for traditional association
body Mab5A10,42 apolipoprotein E4 (ApoE4),30 apo-transfer- models of low molecular weight molecules.52
rin,30 benzylamine,22 β-lactoglobulin,11,33 bovine serum
albumin (BSA),15–20,26,35,39,41,44–47,49,50 chymotrypsin,12 concana-
valin,31 cucurbit[7]uril,34 eosin,48 erythrosine B,48 fluorescein,48 2. Nanoparticles vs. proteins as
fibrinogen,43 human immunoglobulin G (IgG),24,40 human
binding platforms
serum albumin (HSA),14,20,21,30,43,51 Kunitz-type soybean trypsin
inhibitor (STI),15 lysozyme,30,37 phenylalanine,13 phenyl-isothio- When compared to proteins specifically, nanoparticles show
cyanate (PITC),22 recombinant streptococcal protein G (protein differences regarding their binding capabilities, most of them
G),15 S-(thiobenzoyl)thioglycolic acid (PDE-1),23 thiocarbonate derived from their larger surfaces (see example 1, Appendix),
(thiocarbonic acid),25 thiocyanate,28 thioester (1-pyrenebutane- which are pivotal for determining the association capabilities
thioic acid S-butyl ester),25 and macrocyclic polyammonium of nanoparticles and small molecular weight molecules. Some
cations (MCPACs): Me6[14]ane-N4H8 4+,38 tryptophan,13 of the binding qualities of nanoparticles are listed as follows:
trypsin,27 and tyrosine.13 (i) multi-occupation is a common denominator for the
Our overarching goal is to review the fundamental concepts binding of small molecules to nanostructures, which means at
of binding models and how these should be applied to the equilibrium more than one molecule of the substrate will
systems with different size distributions as in the case of nano- be bound per nanoparticle.
materials. In this tutorial review, the reader will learn the basis (ii) Complex competitive binding processes can take place
of binding mechanisms applied to nanomaterials, which will on the nanoparticle surface, which can lead to either coopera-
contribute to building up a rational approach for determining tive or competitive bindings, and
Natalia L. Pacioni obtained her Emilio I. Alarcon received his

chemistry degree (2001) from the B.Sc. in Chemistry from the
Facultad de Ciencias Químicas Universidad de Santiago de
(FCQ), Universidad Nacional de Chile and moved to the
Córdoba (UNC) in Argentina. Pontificia Universidad Catolica
She completed her Ph.D. (2007) de Chile to pursue graduate
at the UNC working in the field studies (MSc/Ph.D., 2009). The
of Supramolecular Analytical same year Dr Alarcon moved to
Chemistry in Dr Veglia’s group. Ottawa with his family for a
After her postdoctoral work postdoctoral position at the
(2008–2011) on applied photo- University of Ottawa. In 2014,
chemistry and metal nano- the University of Ottawa Heart
Natalia L. Pacioni particles with Tito Scaiano, at Emilio I. Alarcon Institute recruited him as an
the University of Ottawa, she Assistant Professor and
returned to Argentina as a CIC researcher for INFIQC-CONICET, Laboratory Director. He has published over 50 articles in peer-
where she is currently an Adjoint researcher. In 2016, she was also reviewed journals, more than 7 book chapters and has edited a
appointed adjoint professor at the Organic Chemistry Department book on nanomaterials for biomedical applications. Dr Alarcon’s
(FCQ-UNC). Her research interests comprise nanomaterials and group works towards the development of nano and biomaterials
supramolecular analytical chemistry. for medical uses.
Table 1 Summary of selected properties of different carriers ( proteins, liposomes, and nanoparticles). Properties shown in the table below are
listed to demonstrate the differences between nanoparticulate materials and other carriers
Property/kind of
carrier Proteinsa Liposomesb Nanoparticlesc
Total binding Fixed [this depends on the number Variable [phospholipid composition Variable [binding capacity depends on
capacity and of binding sites or active surfaces; controls binding extent and the surface the size; surface structure can be
surface surface depends on the amino acids’ structure] manipulated]
composition sequence]
Stability Relatively short half-lives (hours to Variable [depends on the presence of Variable [nanoparticle in vivo half-life
days) [protein oxidation leads to unsaturated bonds and membrane and stability can be varied by changing
reduction in binding] fluidity] the material surface composition]
Size and shape Relatively constant [both size and Fixed and, mostly, spheroids [size/ Both variable [size can be tuned
shape can vary depending on factors geometry depends on the lipid chain (1–100 nm); shapes can be varied]
like ionic strength and pH] length and number of bilayers; shapes
are mostly spherical, because of the
surface tension]
Availability and Limited and costly [depends on the Variable [depends on the nature and Readily available and low cost [bottom-
cost type of protein, protein expression in composition of the liposome, scale-up up routes for preparing nanoparticles
yeast/bacteria is an option but it production is not always possible] are of low cost; surface modification
increases the production cost] could, however, be more expensive]
a
Protein diameter varies depending on the structure, amino acid composition, and ultimately on physiological conditions. Thus, for example in
the case of globular proteins like albumins they can be considered as spherical structures with a 10 nm diameter. b Vesicles’ diameter and shape
vary with the phospholipid content and composition. Sizes are found within 5–200 nm in diameter. Much larger structures are seen when the
formation of supramolecular aggregates composed of multi-layered lipid layers occurs. c Nanostructures are considered as composites that have
one of their dimensions within 1–100 nm in length.
(iii) the formation of a multilayered dynamic regime around Ultimately, the size of the ligand also plays a key role as
the nanoparticle is possible at surface area molecules ⋙ seen from analyzing the association data presented herein,
nanoparticle surface area. where ref. 13, 22, 23, 28, 29 and 48 show the values of Ka
In example 1, we show simple calculations of the total obtained for different small molecules. These association con-
surface area for monodisperse spherical nanoparticles and stant values are considerably lower than the values obtained
compare these numbers with that calculated for a carrier for proteins. This can be explained in terms of either coopera-
protein. The binding capacity for these two systems is analyzed tive binding of the proteins onto the nano surface or by multi-
using the archetypical methylene blue (MB) dye. valency, where in either case the protein will efficiently
To date, the available literature comparing the extent of compete with the nanoparticle surface, to a much greater
association between nanoparticles and carrier proteins is extent than low molecular weight ligands for instance. In
scarce. This is due in part to the significant variability, and section 2.3, we will elaborate more on this aspect.
sometimes limited experimental information available, in Table 2 contains information concerning the size, shape,
research articles, which have dealt with nanomaterial–mole- surface charge (zeta potential), polydispersity, association con-
cule binding parameters. Discrepancies do exist beyond the stant, occupation number and association model or equation
size of the nanoparticle, where different shapes have demon- chosen for determining the association of 80+ nanostructures.
strated to have different, and sometimes opposite trends, for Fig. 1 shows a comparison of the association constant values
the same substrate. For example, Iosin et al. reported a 7 order measured for the binding of molecules to nanoparticles and to
of magnitude difference in the affinity of BSA for spherical human serum albumin (see Table S1† for data on serum
nanogold over rod-shaped nanostructures.17 However, albumin).
Chakraborty et al. reported opposite results, the association of Overall, the magnitude of the extent of association for
nanorods being 16 times larger than the one measured for nanomaterials is considerably larger than that for HSA, with
spherical nanogold.26 average numbers of 7 ± 3 and 4.7 ± 0.8 log Ka (p < 0.001, t-test),
Additionally, nanoparticle size also plays a key role in deter- respectively. Thus, as expected the average association constant
mining the magnitude of the association as seen for example is much higher than that for proteins. However, there is a sig-
in the case of cobalt-iron oxide nanoparticles,44,45 where an nificant overlapping with the values measured for HSA. This
increment of 11 nm, from 5 to 16 nm, increases the extent of differs from what is shown in example 1, regarding the superior
association 24 times for the binding. A similar trend is capacity of nanomaterials to host a much larger number of
reported by Yang et al., where both the association and occu- guest molecules. These discrepancies can be explained if one
pation numbers increased by 33- and 15-fold for the associ- considers some, or all, of the following factors:
ation of α-synuclein with a nanogold with diameters of 20 and i. Nanoparticle colloidal stability is affected.
90 nm, respectively.32 ii. Competitive binding with the capping agent.
Table 2 Selected physical properties and binding parameters for different nanostructures: (macro)molecule systems reported in the literature
Analyst
Zeta Association
potential Polydispersity constant Occupation
Type of NP Substrate Size (nm) Shape (mV) index Binding model (104 M−1) number (n) Ref.
a
Gold Phenyl-isothiocyanate (PITC) ≈6 Spherical — — Benesi–Hildebrand 5.0 — 22
Gold Benzylamine ≈6 Spherical — — Benesi–Hildebrand 5.0a — 22
Gold S-(Thiobenzoyl)thioglycolic acid 16 Spherical — — Linear plot: 230 — 23
(PDE-1) [Ligand]free/
[NP-Ligand] vs.
[Ligand]free
Gold 3-(10-hexyloxyethyl)-3- 53 Cubical +6.4 to — Scatchard 3.1 100 > 180b 29
devinylpyropheophorbide-a cage −26.3
(HPPH)
Gold Human immunoglobulin G (IgG) 24 Spherical — — Scatchard 3300 52b 24
Gold Bovine serum albumin (BSA) 24 Spherical — — Scatchard 1300 90b 15
Gold Recombinant streptococcal 24 Spherical Scatchard 570 500b 15
This journal is © The Royal Society of Chemistry 2017

— —
protein G (protein G)
Gold Kunitz-type soybean trypsin 24 Spherical — — Scatchard 1000 550b 15
inhibitor (STI)
Gold Trypsin 3 Quasi- — — Hill equation 443 — 27
spherical
Gold Trypsin 3 Quasi- — — Modified Stern– 973 — 27
spherical Volmer
Gold Thioester (1-pyrenebutanethioic 12.5 Spherical — — Langmuir 7200 2006 25
acid S-butyl ester)
Gold Thiocarbonate (thiocarbonic acid) — — Langmuir 17 000 2023 25
Gold Concanavalin 32 — — — Y/(1 − Y) = K[L]n/1 + 1.5 × 1011 — 31
K[L]n
Gold Bovine serum albumin (BSA) 8–70 Spherical — — Encinas–Lissi 1.37 × 1010 13 700 16
Gold Bovine serum albumin (BSA) 18 Spherical — — Tedesco 2.34 × 107 1.37c 17
Gold Bovine serum albumin (BSA) 70 × 30 Rods — — Tedesco 5.0 0.38c 17
Gold Bovine serum albumin (BSA) 51 Spherical — — Langmuir modified 0.39 295b 18
Gold Human serum albumin (HSA) 13 Spherical — — Benesi–Hildebrand 33 000 13 (HSA 20
monolayer)
Gold Bovine serum albumin (BSA) 13 Spherical — — Benesi–Hildebrand 34 900 16 (BSA 20
monolayer)
Gold Bovine serum albumin (BSA) 4 Spherical — — Single set of 73.7 0.65e 26
binding sites
Gold α-Synuclein 20 — — — Fluorescence 2.9 × 105 360b 32
quenching
Gold α-Synuclein 90 — — — Fluorescence 9.5 × 106 5300b 32
quenching
Gold Bovine serum albumin (BSA) 20 Rods — — Single set of 1170 0.55e 26
binding sites
Gold Bovine serum albumin (BSA) 10 — ∼−30 (pH — Langmuir 100 2 × 1012 mole- 19
7) cules per cm2
Gold Human serum albumin (HSA) 5–100 Spherical ∼−50/−60 — Single set of 400–4400 9–1215b 21
binding sites
Gold–3,6,9,12- β-Lactoglobulin BLGA 10 — — — Single set of 380 54 33
tetraoxatricosan-1-aminium, binding sites
23-mercapto-N,N,N-trimethyl
Analyst, 2017, 142, 2067–2089 | 2071

Tutorial Review
Table 2 (Contd.)
Zeta Association
Tutorial Review
Gold–3,6,9,12- β-Lactoglobulin BLGB 10 — — — Single set of 210 31 33

tetraoxatricosan-1-aminium, binding sites
23-mercapto-N,N,N-trimethyl
Gold (DHLA) Human serum albumin (HSA) 1.4(TEM)– Cubical −37 — Hill equation 1250 — 30
3.2(DLS) cluster
Gold–dihydrolipoic acid Apo-transferrin 1.4(TEM)– Cubical −37 — Hill equation 1660 — 30
3.2(DLS) cluster
2072 | Analyst, 2017, 142, 2067–2089

Gold–dihydrolipoic acid Lysozyme 1.4(TEM)– Cubical −37 — Hill equation 3.8 — 30
3.2(DLS) cluster
Gold–dihydrolipoic acid Apolipoprotein E4 (ApoE4) 1.4(TEM)– Cubical −37 — Hill equation 500 — 30
3.2(DLS) cluster
b
Gold–4-(dimethylamino) Cucurbit[7]uril 2.5 — — — Single set of 27.6 73.1 34
butyric binding sites
Gold–4-(dimethylamino) Cucurbit[7]uril 2.5 — — — Single set of 13.3 64.6b 34
butyric binding sites
c
Gold–polyacrylate Chymotrypsin 2–8 Quasi- — — Scatchard 400 3 12
spherical
Gold–polyacrylate Chymotrypsin 2–8 Quasi- — — Stern–Volmer 160 — 12
spherical fluorescence
c
Gold–polyacrylate Chymotrypsin 2–8 Quasi- — — Single set of 96 5.5 12
spherical binding sites
Gold–silica and alumina Thiocyanate radical (SCN)2*− 15–25 Quasi- — — Benesi–Hildebrand 0.47 — 28
spherical
a
Silver Tryptophan 20–40 — — — Benesi–Hildebrand 0.254 — 13
Silver Tyrosine 20–40 — — — Benesi–Hildebrand 0.18a — 13
Silver Phenylalanine 20–40 — — — Benesi–Hildebrand 0.314a — 13
Silver Tryptophan 20–40 — — — Fluorescence 15.84 — 13
quenching
Silver Tyrosine 20–40 — — — Fluorescence 4.07 — 13
quenching
Silver Phenylalanine 20–40 — — — Fluorescence 91 200 — 13
quenching
Silver Bovine serum albumin (BSA) 21 Quasi- — — Modified Stern– 0.811 — 15
spherical Volmer
Silver Bovine serum albumin (BSA) 21 Quasi- — — Scatchard 0.783 1.12c 15
spherical
Silver Human serum albumin (HSA) 32 (TEM)– Spherical — — Hill equation 1,408a — 14
42 (DLS)
Silver Human serum albumin (HSA) 32 (TEM)– Spherical — — Linear plot: 3030 — 14
42 (DLS) [Ligand]free/
[NP-Ligand] vs.
[Ligand]free
Silver 1,4-Dimethoxy-2,3- 9–17 nm Quasi- — — Benesi–Hildebrand 0.206 — 36
dibromomethylanthracene-9,10- spherical
dione (DMDBMAD)
Silver Diastereoisomers of tetra 9.62 — — — Benesi–Hildebrand 0.314 — 38
ammonium cations (Me6[14]ane
4HCl)
Silver 9.62 — — — Benesi–Hildebrand 0.338 — 38
Analyst

Table 2 (Contd.)
Zeta Association
Analyst

Silver [32]-Membered octa ammonium 9.62 — — — Benesi–Hildebrand 263 — 38

cation ([32]ane 8HCl)
Silver–para-sulphonatocalix[4] Bovine serum albumin (BSA) — — — — Benesi–Hildebrand 2.4 — 39
arene
Silver–1,3-di-O-phosphonato- Bovine serum albumin (BSA) — — — — Benesi–Hildebrand 0.35 — 39
calix[4]arene
Silver–titanium dioxide Lysozyme 13.6 Non- — — Modified Stern– 1.4 1.03d 37
spherical Volmer
Carbon Human immunoglobulin G 50 Quasi- — — Scatchard 152 0.78c 40
(HIgG) spherical
Copper Bovine serum albumin (BSA) 7.5 Colloidal — — Fluorescence 1288 1c 41
quenching

Polystyrene–COOH Bovine serum albumin (BSA) 60 Quasi- −39.8 — Single set of 24 871b 35
Polystyrene–NH2 Bovine serum albumin (BSA) 58 Quasi- +40.3 — Single set of 4.0 27b 35
Europium(III)-fluoro-max Antibody Mab5A10 107 — — — Scatchard 6.6 × 105 — 42
polystyrene
Aluminum oxide Fibrinogen 312.6 — +20.3 0.559(H2O)– Langmuir 740 — 43
(H2O)– 0.607(PBS)
640.9(PBS)
Aluminum oxide Human serum albumin (HSA) 312.6 — +20.3 0.559(H2O)– Langmuir 44.8 — 43
(H2O)– 0.607(PBS)
640.9(PBS)
Cerium oxide Fibrinogen 200.7 — +26.5 0.372(H2O)– Langmuir 246 — 43
(H2O)– 0.432(PBS)
719.2(PBS)
Cerium oxide Human serum albumin (HSA) 200.7 — +26.5 0.372(H2O)– Langmuir 2530 — 43
(H2O)– 0.432(PBS)
719.2(PBS)
Cobalt–iron oxide–citric acid Bovine serum albumin (BSA) — — — — log[(F0 − F)/(F − Fs)] 4.8 × 108 1.05c 44
vs. log[M]
Cobalt–iron oxide–citric acid Bovine serum albumin (BSA) 5 — — — log[(F0 − F)/(F − Fs)] 8980 1.29c 45
vs. log[M]
Cobalt–iron oxide–citric acid Bovine serum albumin (BSA) 16 — — — log[(F0 − F)/(F − Fs)] 2.12 × 105 0.91c 45
vs. log[M]
Iron oxide β-Lactoglobulin 13–60 — +50 — One-site binding 15.9 — 11
model
Iron oxide–citric acid Bovine serum albumin (BSA) — — — — log[(F0 − F)/(F − Fs)] 21 900 1.96c 44
vs. log[M]
Iron oxide– Bovine serum albumin (BSA) — — — — log[(F0 − F)/(F − Fs)] 31 000 1.2c 46
carboxymethyldextran vs. log[M]
Iron oxide–dextran Bovine serum albumin (BSA) — — — — log[(F0 − F)/(F − Fs)] 33 300 1.85c 44
vs. log[M]
Iron oxide–polyaspartic Bovine serum albumin (BSA) — — — — log[(F0 − F)/(F − Fs)] 19 200 1.45 46
vs. log[M]
Iron oxide–tartrate Bovine serum albumin (BSA) — — — — log[(F0 − F)/(F − Fs)] 30 200 2.42 46
vs. log[M]
Analyst, 2017, 142, 2067–2089 | 2073

Tutorial Review
Ref.
47
43
43
48
48
48
49
50
51
Number of molecules per nanoparticle. c Available number of binding sites. d Binding site for nanoparticles per molecule. e Nanoparticle/protein stoichiometry.
Occupation
number (n)
0.31–2.01c
1.06c
1.88c
—
—
—
—
—
1–483.5×104
Association
0.72 × 103a
1.35 × 106
9.77 × 108
(104 M−1)
constant
10 600
0.088a
1.66a
15.8
1.1a
log[(F0 − F)/(F − Fs)]
Benesi–Hildebrand
Binding model
Stern–Volmer
Stern–Volmer
Stern–Volmer
Fluorescence
Fluorescence
quenching
quenching
Langmuir
Langmuir
vs. log[M]
Fig. 1 Representative association constant distributions for molecules

with nanomaterials ( purple bars) and human serum albumin (HSA, red
bars). Original data are shown in Tables 2 and S1† for nanomaterials and
HSA, respectively.
Polydispersity
0.384(H2O)–
0.384(H2O)–
0.661(PBS)
0.661(PBS)
index
—
—
—
—
—
iii. Underestimation of the nanoparticle concentration, and

iv. Polydispersity is much higher than estimated.
In the following sections, we will briefly discuss the impact
potential
of each of the above-mentioned points including some practi-

(mV)
Zeta
+47
+47
cal examples for illustrating the consequences of each of these

—
—
—
—
—
—
phenomena.
Colloidal
Colloidal
Colloidal
Shape
—
—
—
2.1 Metal clusters as binding platforms

30.6(H2O)–
30.6(H2O)–
735.1(PBS)
735.1(PBS)
Special cases related to the ligand size-dependence are metal

Size (nm)
nanoclusters (NCs), which are characterized by having a small

25–50
1.84
1.84
1.84
2.49
core size (<2 nm), defined molecular structure, discrete elec-

7.5
—
tronic transitions and strong photoluminescence. The litera-

ture on association parameters for ligand-nanoclusters is
scarce, see ref. 30. The combination of these properties
Human serum albumin (HSA)
Human serum albumin (HSA)

Bovine serum albumin (BSA)

together with the use of biomolecules has catapulted the use

of NCs in several biomedical applications, including bio-
imaging, bio-sensing, anti-microbial activity, radiotherapy sen-
sitization, photodynamic therapy, etc. (for detailed reviews, see
Erythrosine B
ref. 54–56). Due to the small size of the metal core in different
Fluorescein
Fibrinogen
nanoparticles, the size of the ligand will modulate the size of

Substrate
the core metal NCs. For example, in the case of Au NCs, it has
Eosin
been shown that when the capping agent is BSA (≈66 kDa) the
final protein–NC complex has around 25 Au atoms (where the
core is imbedded in one protein molecule);57 a similar
Dendrimer-coated magnetite
number of atoms was seen in native lactotransferin

(≈83 kDa).58 However, in the case of lysozyme (≈14 kDa) the
b
Apparent constant.
core has around 12 Au atoms;59 while for polymers like poly

(Contd.)
(amidoamine) (PAMAM; ≈5.0 kDa) and poly(N-vinylpyrroli-

Titanium oxide
Titanium oxide
Zinc monoxide
done) (PVP; ≈10 kDa), 8 and 2–3 (main species) Au atoms

Zinc dioxide
Tin dioxide
Tin dioxide
Tin dioxide
Zinc sulfur
Type of NP
were, respectively, found in the metal core.60,61 Furthermore,

Table 2
the metal core size is highly dependent on the conformation

of the protein.62
a
2.2 Effect of colloidal stability on the nanoparticle binding (zeta potential) and polydispersity.11,19,21,29,30,35,43 Zeta poten-
capacity tial measures to some extent the interaction energy between
In the numbers calculated in example 1, we have omitted particles, which can be related in many cases to the stability of
modifications in the colloidal stability upon the binding of the particles to coagulation, sedimentation and also their flow be-
molecules.63,64 Changes in stability frequently occur after the havior. Thus, if colloidal stabilization is due to the repulsive
replacement of the pre-bound molecules, a.k.a. protecting interaction between particles, zeta potential is a valuable index
agents, from the nanoparticle surface. These can result in a of stability behavior,65 with values close to zero having a larger
net decrease of the available surface for binding (binding probability of flocculation. This phenomenon will decrease the
sites) and consequently, reduce the overall association con- available concentration in solution, thus, resulting in underes-
stant. Fig. 2 shows a representative example of the changes timation of the association constant. Note that for some refer-
in the surface charge of citrate-protected spherical silver ences, sizes and nanoparticle shapes were not reported.
nanoparticles (≈5.0 nm) upon addition of the
2.3 Competitive binding with the capping agent
CSG-LLGDFFRKSKEKIGKEFKRIVQRIKDFLRNLVPRTES-NH 2
(LL37-SH) peptide. 63 Quite frequently, the capping agent that stabilizes the colloidal
From calculating the ratio between the number of peptides nanoparticle solution is seen as a spectator in the binding
and nanoparticles (≈80 nM) available in solution, for the 1, 5 process. However, a simple scheme like the one shown below,
and 10 µM peptide concentrations, the following numbers were Scheme 1, makes it clear that such a non-active behavior is
obtained: 12.5, 62.5, and 125, correspondingly. Thus, consider- instead the result of a much more complex dynamic process.
ing the available surface area of a 5.0 nm nanoparticle as The dynamicity of the capping agent, as well as its role as a
78.5 nm2, see example 1, and since the binding to the surface determining factor for the chemical and biological activities in
of the nanoparticle occurs by anchoring from one of the term- nanomaterials, has been widely overlooked in the literature.
inal amino acids with an area of 1.78 nm2, a maximum number Some more sophisticated examples of these models are found
of 44 molecules of LL37-SH fit onto the surface. This number is for research focused on protein-corona, which certainly played
closer to 62.5 calculated for the 5.0 µM concentration, but a pivotal role in better understanding the formation of areas of
much higher concentrations of the peptide are needed for pro- large molecules around nanoparticles.66 A similar behavior
ducing a net change in the nanoparticle surface charge (125 = has also been reported for low molecular weight molecules
LL37 per nanoparticle). These differences can be linked to steric like methylene blue in the presence of spherical gold nano-
repulsion between the peptide bound to the surface and the particles.67 However, these phenomena are rarely considered
free LL37-SH in solution. Another representative example of in the literature when evaluating association. This becomes
capping agent replacement was also recently reported by using particularly important for nanostructures like silver that are
different pentapeptides, where the sequence CLKRS was able to much more sensitive to capping agent replacement when com-
reduce and modify the stability of citrate-capped AgNPs using pared to gold for instance.68
only micromolar concentrations of this peptide.64 Manipulating In the model depicted in Scheme 1, in principle, one could
the colloidal stability of nanomaterials has proven useful in the think that increasing the concentration of the capping agent
development of advanced techniques for biomedical appli-
cations, see for example ref. 63.
In the data compiled in Table 2, just a small portion of the
articles contained information regarding nanoparticle charge
Scheme 1 Top: Cartoon representation of a multi-step capping agent

binding to a nanoparticle surface. The bound capping agent is shown in
red color and the free, non-bound one in grey. Bottom left:
Representative diagram showing the three different populations of
capping agents once capping of the nanoparticle surface has been
entirely achieved as follows: (1) in surface contact, (2) in the near vicinity
of (1) more organized, and (3) molecules found far from the nano-
Fig. 2 Changes in zeta potential (ζ/mV) for citrate-capped AgNPs, particle. Bottom right: Simplified dynamic model showing capping agent
observed upon addition of increasing concentrations of the LL37-SH exchange within the molecules located in the different layers shown in
peptide. Figure adapted from McLaughlin et al., 2016.63 the left.
will speed up changes in the steps involved in the capping agents of nanostructures than small molecules, thus replacing
exchange between 1 and 2, which does not always hold true. them on the nanoparticle surface. Two different mechanisms
Thus, let us consider that capping agent binding is spon- can explain this assumption. For the first one, we should con-
taneous and diffusion controlled, where layers 1 and 2 are sider that the presence of multiple amino acids in a protein
fully formed at a given concentration of the capping agent. makes it a multivalent structure, thus increasing the number
Thus, due to the proximity of the molecules in region 2 of links to the surface would increase the binding strength
(≈20 nm),67 increasing the capping agent concentration compared to small molecules. Then, if the affinity constant is
should not have a direct effect on the processes between layers larger, a ligand exchange is likely to occur favourably. The
1 and 2. However, since the number of unbound molecules in second explanation is based on the cooperative binding of pro-
solution, region 3, is higher, the rate of exchange between 3 teins, meaning that once a protein binds, the interaction to
and 2, k3B, is faster, and that will lead to a more rapid dynamic the surface for another protein is facilitated successively,
exchange between the layers. In practice, the extent to which resulting also in a larger Ka.69 However, more research on this
this increase in capping agent concentration can affect the aspect is still needed in order to be conclusive on which or if
binding of another molecule to the nanoparticle surface is both mechanisms are operating.
hard to quantify. However, qualitatively, it was recently Canoa et al. found that the molecular weight (MW) of the
reported, see Fig. 3, that increasing the concentration of citrate protein dictates the extent of association of proteins with tita-
15 times does displace a silver specific peptide CLKRS from nium oxide (TiO2), and aluminum oxide (Al2O3).43 A similar
spherical silver nanoparticles.64 trend was reported by Shang et al. who reported a trend for the
The association of the CLKRS peptide with the surface of association as follows: Apo-transferrin > ApoE4 > lysozyme to
spherical silver nanoparticles leads to a broadening of the dihydrolipoic acid capped nanogold.30 Furthermore, similar
surface plasmon band (SPB) of silver, which is reduced as the association constants have been reported for BSA and HSA to
citrate concentration in solutions increases, see Fig. 3. Thus, the nanogold homology.20 However, there is no link between
increasing citrate concentration does displace a pre-bound the MW and affinity for the association of other globular pro-
CLKRS peptide from the silver surface. Therefore, one can teins with nanogold.15 Furthermore, two other proteins, BLGA
expect that the extent of association of this CLKRS will indeed and BLGB, show differences in affinity to 3,6,9,12-tetraoxa-
depend on the citrate concentration in solution. tricosan-1-aminium and 23-mercapto-N,N,N-trimethyl capped
In two independent research studies by Tedesco et al. and nanogold.33
Macarof et al., the authors reported similar association con- As a summary, in this section, we illustrated that the
stant values for the binding of BSA to iron oxide nanoparticles surface composition, particularly the capping agent concen-
capped with different agents like carboxymethyldextran, tration and nature, should also be considered and reported
dextran, polyaspartic, tartrate and citric acid.44,46 Similar find- when measuring the association constants and affinities of
ings were reported by Tonga et al. and Taura et al. for the molecules towards nanostructures. Failing to do this has con-
binding of curcubit[7]uril to AuNP and BSA to AgNP, respect- tributed to the variability in the extent of association reported
ively.34,39 Although these research studies used a fixed concen- for small and large molecules to nanoparticles that are cur-
tration of the capping agent, their findings seem to indicate rently found in the literature.
that proteins have a much higher capacity to act as capping
2.4 Pitfalls in determining the nanoparticle concentration
Estimation of the actual nanoparticle concentration remains a
complicated task for materials scientists. This still implies the
use of assumptions that ultimately represents the primary
source of error (relative standard deviation, RSD, ca. 10%). To
determine with accuracy the concentration of a given nano-
particle in solution, the following pieces of information are
required. (i) Starting concentration of the nanoparticle precur-
sor. (ii) Redox extent of the nanoparticle component. (iii)
Packing or arrangement of the single unit elements in the
nanoparticle, and (iv) nanoparticle shape. Of all these, the
easier value to access is (i), while (ii) can be quantified after
nanoparticle precipitation followed by measurements of the
ion/nanoparticle concentration in the supernatant. Points (iii)
and (iv) go along as both are calculable by using transmission
electron microscopy (TEM) techniques, see Fig. 4.
Example 2 is a representative depiction of one of the many
Fig. 3 Effect of adding sodium citrate to a colloidal solution of CLKRS
2.0 µM and citrate-capped AgNPs. Citrate concentrations are expressed
other approaches that have been proposed for determining
in folds, 1× being equal to 1.0 mM. All measurements were carried out at nanoparticle concentration in solution. However, this only
room temperature. Figure adapted with permission from Poblete et al.64 operates for materials with crystalline and a periodic structure;
effect on the resulting association constant for a given mole-

cule–nanomaterial system. In the following sections, by using
spherical AgNPs as a model, we will didactically demonstrate
the relevance of size polydispersity on the total surface area for
a given solution of nanoparticles. Example 4 contains some
different discrete size distributions for spherical silver nano-
particles all prepared using 0.2 mM of silver. In our calcu-
lations, we have assumed total silver reduction during
Fig. 4 Representative transmission electron microscopy (TEM) images synthesis.
of reshaped silver nanoparticles prepared according to the literature. Example 4 clearly illustrates the impact of size disparity
The inset corresponds to a high-resolution of the nanorod shown in the for a given population of spherical nanoparticles, with a
figure. Original data reported in ref. 70. mean size distribution centered at 10 nm. Thus, for example,
one can see how the total surface area, which ultimately con-
trols the binding capacity of a nanomaterial, can vary up to
see ref. 71, for a review on amorphous gold structures. When 10 times when smaller nanosurfaces are available for
the material is non-crystalline a.k.a., amorphous, using calcu- binding. Although in our example we have not considered
lations like the one shown in example 2 is not possible. Other continuous size distribution, it clearly shows how slight
different calculation procedures are also found in the litera- changes in the sample size dispersity can have a tremendous
ture. We will briefly look into some of them, and we hope they impact on the binding capacity of a given solution of
will serve researchers who are working in the field of nanoparticles.
nanomaterials. A more elaborate picture that encompasses the size distri-
There is more than a single way of calculating nanoparticle bution with the total number of nanoparticles, and ultimately
concentration, and every one of them will render different the total surface area, can be obtained through the nano-
final concentrations. In the following sections, we will briefly particle size distribution function. Such a function can be rep-
use four different methods for calculating the concentration of resented by a normalized Gaussian distribution (g: R+ → R+),
spherical nanogold, see example 3. supported on the diameter interval (a, b). More precisely: (i)
Table 3 shows a comparison of the obtained results for g(t ) = 0 if t ∉ (a, b), (ii) g(t ) > 0 if t ∈ (a, b), (iii) g(c) = 1 where
stock solutions of nanogold, calculated using the different c = (a + b)/2, (vi) g(c − h) = g(c + h) for every 0 < h < c − a and
methods described in example 3. It is interesting to note that ða
method 1 gives the lowest associated errors when compared to
gðtÞdt ¼ 1
method 3. However, as a summary, it becomes evident that the b
variability of nanoparticle concentration is directly linked to
the methodology used for calculating the nanogold The above allows the following analysis. Given a ≤ d < b and
concentration. h ≈ 0 such that d + h ≤ b, the probability of finding a nano-
particle with a diameter d is given by:
2.5 Polydispersity is much higher than estimated ð dþh
Size-dispersity for a given population of nanoparticles with the gðtÞdt gðdÞh
d
same geometry does have a considerable impact on the total
surface area available in solution. These differences are typi- Thus, there is a g(d )h percentage of nanoparticles with dia-
cally neglected as it is assumed that these would have little meter, d, of the total number of nanoparticles in a sample
with distribution g.
We will assume that the total concentration of the individ-
ual components for each nanoparticle is known. Thus, we can
Table 3 Comparison of nominal concentrations of nanogold in solu- denote such a concentration as N0 and let g0(t ) = N0g(t ). From
tion estimated using different methods described in example 3 the preceding, the number of nanoparticles of diameter d is
equal to g0(d )h. Assuming that there must be a function
Concentration of stock solutiona/nM
A: R → R that gives to each d the value A(d ), which corresponds
Type Method Method Method Method with the surface area of a nanoparticle of diameter d, the total
of NP Size/nm 1 2b 3c 4d surface area of nanoparticles of diameter d is equal to A(d )
Nanogold f 12.8 (50 ± 3)e 52 23 ± 17 57 g0(d )h. Considering that Pn = {a = d0 < d1 < ⋯ < dn < b ≤ dn+1} is
(37 ± 2)e 38 17 ± 12 42 a partition such that dj+1 − dj = 1/n, we conclude that the total
(50 ± 2)e 51 22 ± 16 57 surface area relative to the partition Pn is given by the
a
Values of different stock solutions. b Error in the ε calculation is extre- Riemann summation:
mely high (>100%). c Error in the ε calculation is around 73%. d Error
in the ε calculation is extremely high (>100%). e Calculated using the X
n X
n
Aðdi Þg0 ðdi Þ
ε520 = (0.24 ± 0.01) nM−1 cm−1. f NP size: 12.8 nm; RSD: 4.2. All uncer- SðPn Þ ¼ Aðdi Þg0 ðdi Þðdiþ1 di Þ ¼
n
tainties were estimated using error propagation. i¼0 i¼0
Finally, and taking limits in the Riemann sum, the total Table 2 include occupation numbers of 30 to 5000+ proteins
surface area is given by: per nanoparticle.15,24,29,32,33,35 Protein corona formation
ðb ðb around the nanoparticle is certainly a plausible explanation,66
lim SðPn Þ ¼ AðtÞg0 ðtÞdt ¼ N0 AðtÞgðtÞdt ¼: Atotal for at least a small fraction of the 30+ molecules, which can be
n!1 a a also combined with irreversible protein aggregation around
One can further elaborate this expression considering a the nanoparticle.
probabilistic approach. Let Xg be a random variable over R
where the probability density function is a Gaussian function
g. Thus, the distribution function of Xg is given by: 3. Determining the association
ðx constant of ligands with nanoparticles:
FXg ðxÞ ¼ P½Xg x ¼ gðtÞdt;
1 pros and cons of the current
where P is a probability on R. The expected value of the methodologies
random variable Xg is given by
In the previous section, we have shown the main differences in
ðx ðb
the association behavior, which exist between nanoparticles
E½Xg ¼ gðtÞdt ¼ gðtÞdt
1 a versus other systems. Also, we have discussed several factors
affecting association constant values like: the under/overesti-
With the function A given, we can define a new random
mated concentration of nanoparticles and occupation number
variable Y = A(Xg); the probability density function of Y is
(n), influence of polydispersity on the available surface area,
unknown; however the expected value of Y can be calculated
colloidal stability and competitive binding.
by the equality:
Now, let us discuss the pros and cons of some of the most
ð þ1 ðb used methodologies to measure “concentrations” in order to
E½Y ¼ E½AðXg Þ ¼ AðtÞgðtÞdt ¼ AðtÞgðtÞdt determine the binding/association constant (Ka). First, we
1 a
need to recall the association constant, a thermodynamic para-
which is sometimes called the law of the unconscious statis- meter which reflects the strength of interaction between a host
tician. Therefore, we have or receptor (S) and a guest or ligand (L) at equilibrium:
N 0 E½AðX g Þ ¼ Atotal K ½Sm Ln
mS þ nL Ð Sm Ln ; K¼
½Sm ½Ln
The derivation of equations done in the previous para-
graphs was specifically calculated for the total surface area (A). For nanoparticles, the number of substrates (m) is usually
This can also be extended for calculating the total number of equal to 1, but the number of ligands (n) is often much larger.
atoms in a given nanoparticle as follows. Let F: R → R be a Thus, we should consider dealing with multiple equilibria
function that gives to each d the value F(d ) which corresponds systems in the best scenario (binding phenomena being revers-
with the number of atoms of a nanoparticle of diameter, d, ible). Obtaining an accurate Ka value, in any case, involves the
with X0 being the total number of atoms in the sample as knowledge of the complex stoichiometry. When this parameter
follows: is not known or in the case of multiple equilibria existing,
then we are likely only to determine an overall K value (βmn =
N 0 E½FðX g Þ ¼ X 0
ΠKi) for the system.81,82
then the total concentration in the sample can be expressed The next step is to find a mathematical model that allows
as: calculating the association constant through the relationships
between a measurable physical or chemical property and the
X0
N0 ¼ concentrations (at equilibrium) of the compound, the nano-
E FðXg Þ
particles or the ligand (Table 4). Choosing the right method-
Another parameter closely related to the association of ology to quantify any of these components is critical in order
molecules with nanoparticulated surfaces is the occupation to obtain an accurate and precise K value. In the following
number, usually represented as ‘n’. This parameter relates the paragraphs, we will discuss typical cases for the most frequent
number of molecules adsorbed in/interacting with a given methodologies found in the literature. Note that most of these
surface. As n is derived from a measurement, it is directly methods are based on titration experiments in which either
linked to the method used for its quantification. However, due the nanoparticles or the ligand concentration is kept constant
to the limited surface on the nanoparticle surface, one can and the other is varied.
expect that the number of molecules bound per nanoparticle
will be much higher when using low molecular weight mole- 3.1 UV-visible spectroscopy
cules than compared to macromolecules. In all the discussions This technique requires that either the nanoparticle or the
to date, we have only considered the formation of a monolayer, ligand absorbs light in the UV-visible region. So, nanoparticles
full surface saturation. Note that some of the references in presenting a plasmon band and ligands with chromophores in
Table 4 Representative examples of the use of methodologies associated with different models for calculating association constants of molecules
or macromolecules with nanoparticles
Molecules or (bio)
Technique Equation/model useda macromolecules Ref.
UV-Vis spectroscopy [NP] = constant; [Ligand] = variable Phenyl isothiocyanate (PITC) 22

Benesi–Hildebrand approach Benzylamine
1 1 1
¼ þ
Aobs A0 AC A0 Kapp ðAC A0 Þ½Ligand
Linear plot: [Ligand]free/[NP-Ligand] vs. [Ligand]free Phenyldithioesters (PDE) 23
From slope and intercept Kads (adsorption constant) and Γmax (complex at Tryptophan 13
saturation) are determined
[NP] = variable; [Ligand] = constant Tyrosine
Benesi–Hildebrand approach Phenylalanine
1 1 1
¼ þ
Aobs A0 AC A0 Kapp ðAC A0 Þ½NP
Scatchard approach 3-(10-Hexyloxyethyl)-3- 29
½Ligandfree devinylpyropheophorbide-α
½Ligandbound ¼ n (HPPH)
Kd þ ½Ligandfree
[Ligand]bound = [Ligand]total − [Ligand]free.
Kd and n are the dissociation constant and number of binding sites,
respectively.
Fluorescence Residual fluorescence of the reaction solution after separation from Human immunoglobulin G 24
nanoparticles by centrifugation. Fluorescence of tryptophan at 350 nm, (HIgG)
Scatchard model
½Ligandbound n ½Ligandbound
¼ Bovine serum albumin (BSA)
½Ligandfree Kd Protein G
Kunitz-type soybean trypsin
inhibitor (STI)
From quenching of ligand fluorescence by NP (Q) log[(F0 − F)/F] = log K Tryptophan, tyrosine and 13,
+ n log[Q], where n is the number of occupied binding sites phenylalanine 27
Lysozyme and
Trypsin 37
Modified Stern–Volmer equation BSA 15
F0 1 1
¼ þ
ΔF fKa ½Q f
ΔF = F0 − F at a quencher concentration [Q], f is the fraction of accessible
fluorescence, and Ka is the associative binding constants for the
quencher–acceptor system.
F Fmin
log ¼ m log K þ n log½Q Trypsin 27
F
m is the Hill’s coefficient (which value indicates cooperativity or
anticooperativity interaction) and n is the number of occupied binding
sites
Scatchard approach HIgG 40
r
¼ ðn rÞK BSA 15
½Ligandfree
r is the moles of ligands bound per mole of protein, n is the binding site
multiplicity per class of binding sites, and K is the equilibrium binding
constant.
Enhancement of nanocluster fluorescence upon adsorption of protein. Human serum albumin (HSA), 30
Hill equation. KD is the dissociation constant and n is the Hill coefficient. transferrin, lysozyme, and
Fmax F0 apolipoprotein E4 (ApoE4) 84
F ¼ F0 þ
KD n
1þ
½HSA
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
Rh ð½BSAÞ ¼ Rh ð0Þu
VBSA N
Plasmon scattering 3 1 þ
u BSA 18
correlation t VNP KD n
1þ
spectroscopy ½BSA
VBSA and VNP are the volumes of the BSA and the AuNP, respectively, N is DNA bases
the number of proteins bound, KD is the dissociation constant, and n is
the Hill coefficient
Table 4 (Contd.)
Molecules or (bio)
Technique Equation/model useda macromolecules Ref.
Isothermal titration Integrated calorimetric response vs. titrant volume. Interaction was Aspartic and lysine 87
calorimetry (ITC) corroborate but no constants were determined and
88
Single-site binding model. Q is the total released heat and a is a constant. HSA 21
½complex Q=a
K¼ ¼ β-Lactoglobulin (BLG) 11
½sitefree ½proteinfree ð½sitetot Q=aÞð½proteintot Q=aÞ
solving for Q:

a 1
Q¼ ½proteintot þ n½NPtot þ
2 K
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi9
= BSA, BLG-A and BLG-B 33
1 2
½proteintot þ n½NPtot þ 4n½proteintot ½NPtot
K ;
Multiple non-interacting binding sites (n) model. Q is the total released heat Suwannee River humic acids 86
(SRHAs)
dQ 1 ½L 1
¼ ΔH 1 þ 1
dnL 2 n½S nK½S
3
2 1
½L 1 4½L 2 5
1þ þ
n½S nK½S n½S
Langmuir isotherm BSA 19
Quartz crystal Γ max KL ½Sfree
microbalance Γ BSA ¼
a þ KL ½Sfree
Γ represents the surface coverage and KL is the binding constant.
1
Brownian Motion Θ ¼ Θmax
HSA 14
Nanoparticle Sizer KD n
1þ
(BMNS) ½HSA
Θ is the surface coverage, n is the Hill’s coefficient and [HSA] is the
protein concentration
½P0 1
Circular dichroism 1 ¼ ½S HSA 79
P KD
[P0] is the total protein amount, [P] is the free protein concentration and
[S] is the concentration of NP surface sites.
SPR using the Fibrinogen-NP: Langmuir model describing a 1 : 1 binding stoichiometry HSA 43
ProteOn™ XPR36 (A + B=AB). Fibrinogen
Protein Interaction HSA–NP: two states conformational change model
Array System (Bio-Rad) (A + B = AB = (AB)a)
Cyclic voltammetry Molar-ratio method. Cucurbituril-7 (CB) and 92
Binary systems (NP@CB and NP@MB) Methylene blue (MB)
NPm + xL ⇔ MmLx

1 αML 1
log½cL xαML cNP ¼ log log K
x 1 αML x
cL and cNP are ligand concentration and nanoparticle concentration,
respectively; αML is the molar fraction of complex and x is the stoichio-
metric coefficient.
Ternary systems (NP@CB@MB = MmLyBz)
For x = y
αMLB
Km;x;z ¼
ð1 αMLB ÞðcB ðz=mÞαMLB cNP Þz
αMLB, cB, are the molar fraction of the ternary complex and the concen-
tration of the third component. z and m are the stoichiometric coeffi-
cients for B and NP, respectively.
Localized Surface Langmuir adsorption isotherm. Lipopolysaccharides (LPS) 91
Plasmon Resonance Δλ K½LPS
(LSPR) refractometric ¼
sensing Δλmax 1 þ K½LPS
a
Parameters and variables have been named as in the corresponding reference.
their chemical structure are suitable to be quantified by UV- the occupied binding sites. In a similar manner, Zhang et al.40
visible spectroscopy. Second, there has to be a significant determined the binding constant of carbon nanoparticles and
difference in absorption between the bound and free com- human immunoglobulin G (HIgG).
ponents to determine the association constant (εb ≠ εf ). Third, A different approach involving the enhancement of a fluo-
if both nanoparticles and ligands absorb light, then choosing rescence signal of gold nanoclusters upon addition of HSA
a wavelength that avoids spectra overlap would be rec- (and other proteins) was used to determine the affinity con-
ommended. Otherwise, the absorption values must be cor- stant. Shang et al.30,84 employed Hill’s equation (Table 4) to
rected before using any suitable model to calculate the Ka. estimate the Ka value and attribute an anti-cooperative process
For example, following the surface plasmon absorption of (Hill’s coefficient < 1).
gold nanoparticles upon the addition of increasing amounts Using fluorescence techniques requires additional con-
of phenyl isothiocyanate (PITC), Thomas et al.22 determined siderations to those mentioned above for the case of UV-visible
an apparent K value, Kapp ≅ 104 M−1 (Table 2) applying the spectroscopy. For instance, it is critical to avoid quencher
Benesi–Hildebrand (B–H) approach (Table 4) to analyze the absorption at both the excitation and the emission wavelength
absorption data. In a similar fashion, Roy et al.13 estimated of the fluorophore. Otherwise, these interferences will produce
the Kapp for AgNP and amino acids following the UV-visible excitation/emission inner filter effects, also producing attenu-
variation in the region of tryptophan, tyrosine and phenyl- ation of the fluorophore emission signal, and a correction will
alanine upon the addition of nanoparticles. However, this be required to account for the ‘real’ quenching. However, we
work did not mention the chosen wavelength or if any correc- should have in mind that any of the correction formulas are
tion was applied. In a different manner, Blakey et al.23 used approximate.85 In addition, as now the fluorescence is used to
the π–π* absorption band of a phenyldithioesther (PDE) and estimate concentrations, sample absorbance at the excitation
the surface plasmon band variations of AuNPs in the presence wavelength should not exceed 0.05 to avoid deviations larger
of PDE to calculate free ligand, and the complex concentration than 5% from linearity.
at saturation and estimated free sites on the nanoparticle.
Thus, using a linear model (Table 4, equation not specified) 3.3 Isothermal titration calorimetry (ITC)
that resembles a y-reciprocal plot, they calculated the equili-
This technique measures differences in heat between a refer-
brium constant (Table 2).
ence cell and a sample cell (where the titration is taking place).
Although linear regression methods like Benesi–
Initially, both cells are at the same temperature, but when
Hilderbrand and Scatchard are out-dated and poorly rec-
titration begins the interaction between the ligand and recep-
ommended due to the errors they introduce,81 they are still
tor will produce an exchange of heat, which is then detected.
often used as seen in the reported association constants
Thus, using the appropriate model, all the thermodynamic
shown in Table 2. Particularly, the B–H model is valid for 1 : 1
parameters (Gibbs free energy, enthalpy, and entropy) besides
complexes, though the model is not as sensitive to deviations
the binding constant and stoichiometry ratio can be calculated
due to multiple equilibria as often believed. Also, B–H usually
in only one experiment.86
requires a second assumption: [L]0 ([S]0) equals unbound/free
Note that ITC is not used to determine concentrations, but
[L] ([S]). The first one will introduce error to the constant for
to directly calculate thermodynamic and reaction parameters
the case of multiple equilibria, while the second assumption
needs pseudo-first order conditions to minimize the error.
Besides, Ka is obtained from a ratio between the intercept and
slope, increasing the error magnitude. Nonlinear regression
will give more reliable results. Note that all the mentioned
models used to estimate the association parameters could be
derived from the data obtained by different techniques.
3.2 Fluorescence spectroscopy

Changes in fluorescence emission due to the interaction
between any molecules and nanoparticles can be monitored to
determine the association constant. These variations can
involve the quenching or enhancement of fluorescence
signals.12,13,15,27,30,32,37,40,41,44–48,83,84
In the case of static quenching, a Stern–Volmer equation
can be used to determine the value of Ka. For instance, Jiang
et al.15 analyzed the quenching of BSA by a AgNP using a Fig. 5 Comparison of adsorption isotherms for the interaction of AgNP
and BSA fitted using a modified Langmuir model with Hill coefficients of
modified Stern–Volmer model to calculate the Ka (Table 4).
0.4 (blue line) or 1 (black line). Experimental data (red squares) of hydro-
The obtained value was comparable to the same parameter dynamics radius are measured using plasmon scattering correlation
determined using the Scatchard model. The main difference is spectroscopy. Reproduced with permission from Dominguez-Medina
that the last one also allows obtaining an estimated value of et al.18 ©2012 American Chemical Society.
from the binding isotherm plotted as exchanged heat against hydrodynamic radius (Rh) through detecting variations in the
the molar ratio of reactants. This fact is important to consider, Brownian diffusion of AuNPs upon the addition of BSA. Then,
as it makes it essential to know the nanoparticle concentration using a modified Langmuir model (Table 4) they were able to
accurately to establish a good molar ratio. calculate the dissociation constant and attribute an anti-coop-
For example, Sastry et al.87,88 plotted the binding isotherm erative interaction. The revised model includes the Hill coeffi-
against the total volume of the ligand added. From these data, cient (n) to measure the cooperativity factor, whose value has
trends of binding stability for DNA bases or amino acids with to be given until the best fit is reached. However, the differ-
gold nanoparticles could be obtained, but no association con- ences seen when n = 1 versus n = 0.4 (best fit) are not as signifi-
stant was determined. The authors considered that no accurate cant if error bars in experimental data are considered, see
information about the number of moles of surface atoms was Fig. 5.
known, then the molar ratio of the ligand to nanoparticle Using a laboratory made Brownian motion nanoparticle
could not be precisely determined. sizer, Gebauer et al.14 measured the surface coverage of HSA
Most recently, other authors have ventured into determin- on AgNPs. This variable was related through a Hill equation
ing the actual value of Ka. The central assumption they made (Table 4) with the protein concentration to calculate the dis-
is that the binding process follows a multiple non-interacting sociation constant. In the same work, they also did an inde-
site model, which is equivalent to the single set of binding pendent experiment using circular dichroism to quantify the
sites fitting model any ITC software brings incorporated in it binding. A good point they brought up is that protein–nano-
(see Table 4). This model works well as far as all the binding particle interactions are likely to be a non-equilibrium process,
sites are identical (the same Ka and the same enthalpy change) and so association constants have to be examined with
and the binding behavior is non-cooperative. For instance, caution.
using this model, Goy-López et al. calculated the Ka for HSA to Brewer et al.19 used a quartz crystal microbalance with dissi-
AuNPs of different sizes,21 Qin et al. estimated the association pation to monitor the frequency change upon the binding of
of β-lactoglobulin (BLG) with polymer-grafted magnetic nano- BSA to gold nanoparticles and surfaces. From these measure-
particles11 and Chen et al. did the same for BSA and two types ments, the surface coverage can be determined, then the
of BLG.33 Using a slightly different equation, Loosli et al.86 binding constant was calculated using a Langmuir isotherm
(Table 4) determined the binding affinity of TiO2 nanoparticles model (Table 4). Again, caution is recommended when report-
to River humic acids. ing the Ka value as a non-equilibrium situation can take place
Although these examples assume non-cooperativity, which due to spreading, resulting in a decrease of the apparent
would mean the consecutive ligand bindings are not affected binding constants.
by the previous ones, the ITC technique can also help, some- Local Surface Plasmon Resonance (LSPR) refractometric
times, to determine if the sites are interacting (sequential sensing was used to determine the peak shift of the plasmon
binding site model). Also, this model can also be useful for of gold nanorods upon interaction with lipopolysaccharides
the case of non-identical sites. (LPS).91 Plots of the relative peak shift versus LPS concentration
The ITC technique requires a sound knowledge of the con- were also fitted using a Langmuir isotherm model (Table 4) to
centration of the reactants, which we have already mentioned estimate the Ka values.
is not accurate for nanomaterials, and assumptions on the In another case, kinetics experiments based on SPR
type of binding site (identical, non-interacting, etc.). In measurements performed using a ProteOn™ XPR36 Protein
addition, even for the case of one set of sites, it is important to Interaction Array System (Bio-Rad) allowed the determination
consider in the analysis if the ligand is the titrant or the one of association and dissociation rate constants, which ratio is
being titrated, as this can lead to different results. Moreover, equal to the binding constant. This technique was used to
Stoll et al.86 found that when the nanoparticles are the titrant, analyze the reversible interaction of fibrinogen and HSA to
the process is driven mainly by entropy whereas in the oppo- metal oxide nanoparticles.43 The parameters for fibrinogen
site case it is driven by enthalpy, though the Gibbs free energy were obtained from the binding sensorgrams using the
values were similar. Langmuir 1 : 1 isotherm model whereas a two-state confor-
In order to get accurate ITC results it is very important to mation change model was used to fit HSA data (Table 4,
account for any non-specific effect on heat changes, such as equations not specified).
the heat of dilution of the titrant (or the titrate) into the In a different approach, Doménech-Carbó et al.92 employed
medium, by running the appropriate control experiments (e.g., cyclic voltammetry and the generalized molar ratio method
injecting the ligand into the buffer). Although, this is a very (Table 4) to determine the apparent binding constants and
common experimental procedure for any ITC user, only a few complex stoichiometry of binary and ternary systems contain-
articles we examined clearly indicated having done the men- ing AuNPs. Measurements were based on the current for the
tioned control experiment.89,90 reduction of dissolved oxygen at a Pt microelectrode in contact
with a colloidal gold solution containing different ligand con-
3.4 Other techniques centrations (cucurbituril-7 and MB).
Dominguez-Medina et al.18 used Plasmon Scattering Table 5 contains a summary for the main pros and cons of
Correlation (PSC) spectroscopy to monitor changes in the using the different association models for evaluating the
Table 5 Summary of the main advantages and limitations for some selected methodologies most commonly used for determining the association
of molecules with nanostructures
Association model
representation Equationa Advantages Limitations
K11 ½S
Nonlinear y¼ No approximations are needed Requires initial estimates of all parameters
1 þ K11 ½S Errors are minimized if a good Accuracy and precision are limited
number of points are obtained (maximized error) if complexed ligand
A nonlinear expression can be fraction is out of the range 0.2 to 0.8, being
obtained for complexes of higher ideal to reach the 75% of complete saturation
1 1 order
Benesi–Hildebrand ¼1þ Linear relationship Out-dated method
Or double-reciprocal y K11 ½S It can be helpful to estimate Only valid for 1 : 1 stoichiometry
initial parameters for nonlinear Associated error for the K value is larger than
regression for nonlinear regression
Accuracy and precision are limited
(maximized error) if complexed ligand
fraction is out of the range 0.2 to 0.8, being
ideal to reach the 75% of complete saturation
It can lead to illusory solutions by seeking
y local minima
Scatchard ¼ K11 y þ K11 Linear relationship Out-dated method
½S It can be helpful to estimate Associated error for the K value is larger than
initial parameters for nonlinear for nonlinear regression
regression
Deviations from linearity are Accuracy and precision are limited
related to cooperative binding (maximized error) if complexed ligand
phenomena fraction is out of the range 0.2 to 0.8, being
Hill y Deviations of n from 1 are related Accuracy and precision are limited
log ¼ log½S n log K11 ; where
1y to cooperative (positive or (maximized error) if complexed ligand
n is the Hill’s coefficient. negative) binding phenomena fraction is out of the range 0.2 to 0.8, being
Stern–Volmerb y = 1 + KSV[S] Linear relationship Concomitance with dynamics quenching can
Straightforward if static lead to wrong conclusions
quenching is confirmed
Modifications can allow
determining binding sites
number
a
Considering the simplest case (1 : 1 complex stoichiometry) equations are derived for the fraction of unbound ( f ) ligand (guest); y = f11 = [SL]/
[L]t. A similar equation can also be obtained for the substrate (nanoparticles). b Model derived specifically from fluorescence measurements, KSV
is equivalent to K11 when quenching is static.
association of low molecular weight molecules with nano- changes in the most common techniques for determining the
structures. Particular emphasis must be placed on the associ- association (fluorescence or absorbance) of the molecule that
ated errors derived from using data treatment for the resulting changes upon association with the nanostructure.
association constant values. For a deeper analysis on the deri- Scheme 2 aims to provide a straightforward and simplified
vation of association models we encourage reading the review list of steps for expediting an accurate determination of the
by P. Thordarson.81 association constant of molecules with nanoparticles. Note
how we have not included factors like polydispersity in this
scheme. Scheme 3 shows another simplified diagram for the
determination of association constants of molecules with
4. General considerations for nanoparticles but following changes in the nanoparticle
determining the association of surface plasmon band or surface charge of the nanostructure
molecules with nanoparticles upon adding increasing concentrations of the ligand
(molecule).
In order to decide what protocol/steps should be taken for eval- From what is shown in Schemes 2 and 3, it becomes
uating the association constant/magnitude of molecules with evident that evaluation of the association of non-fluorescent,
nanoparticles one should consider the properties of both the poorly absorbing molecules, or those with a similar surface
nanoparticles and molecules. Scheme 2 contains a simplified charge to the parental surface charge of the nanoparticle is
diagram of some recommended steps one should follow for rather limited. However, in these particular cases, physical sep-
determining the association of molecules when monitoring aration of the nanostructures and evaluation of the free frac-
Scheme 2 Simplified diagram showing recommended steps for evaluating the association of molecules with nanostructures using fluorescence or
absorption changes of the substrate (molecule).
Scheme 3 Simplified diagram showing recommended steps for evaluating the association of molecules with nanostructures following changes on
the surface plasmon band or surface charge for assessing the association of a given substrate (molecule) with the nanostructures.
tion of the molecules in solution are the only reliable alterna- 5. Outlook and future
tives for an accurate assessment of the association constants.
However, in cases where the portion of the bonded compound The knowledge gained from association models and tech-
is too small to be detected, more sensitive techniques like ITC niques aimed at their evaluation has paved the way for
micro-calorimetry are better suited for providing reliable infor- advances in the nanosciences field. However, when compared
mation on the association constants. to proteins or other supramolecular systems, nanoparticles
have some differences including the formation of organized

layers of ligands on surfaces, which is a consequence of multi-
occupation. Such a phenomenon is responsible for the
changes in nanoparticle stability and surface composition,
which ultimately dictates the activity and performance of a
given nanoparticle in a closed environment (Scheme 1).
Despite the relevance of gathering accurate association para-
meters for molecules with nanostructures, to this day some
parameters closely related to the preparation of the nano-
particle (size, polydispersity, capping agent) have interfered In (a) binding takes place from the longitudinal side, while for
with a precise determination of the binding parameters. (b) association occurs from the transversal side. Thus, one can
These, and other, parameters have in fact a tremendous calculate the maximum number of MB molecules per nano-
impact on, for example, the number of nanostructures per particle, which renders 242 and 990 particles, for arrange-
volume, total surface area and affinity for ligands. ments (a) and (b) in that order. Similar calculations can be
Furthermore, having a depository platform for ‘raw’ data used done for the protein, giving 2.42 and 9.90 molecules of MB for
in the calculation of association constants for nanostructures (a) and (b), respectively. Next, we can calculate the total surface
will allow a more precise comparison of the extent of binding, area available for both platforms, by multiplying the original
of the same pair of nanoparticle–molecule, calculated by surface by the total number of nanoparticles (100 nM; 6.022 ×
different methodologies. 1016 nanoparticles ( proteins) per L), which are 1.89 × 1019 nm2
Although no equation will be able to account for all the and 1.89 × 1017 nm2 for nanoparticles and protein, corre-
critical parameters entirely, anyone interested in determining spondingly. Thus, one can calculate the maximum number of
association constants of molecules with nanoparticles must MB molecules per L for each configuration as:
incorporate and account for techniques that are suitable for
each particular case (see Schemes 2 and 3). Furthermore, one
MB (nM)
should also account for the batch-to-batch variability, which
implies the evaluation of the binding parameters using (a) (b)
different batches of the same kind of nanoparticle.
Nanoparticle 24 220 99 039
Protein 242 990
Note that these numbers correspond to a maximum

number of MB per surface area, and they do not consider, par-
ticularly for the case of the protein, phenomena like competi-
Example 1. Nanoparticle vs. protein tive binding. Furthermore, for the pocket binding, we have not
binding capabilities accounted that binding occurs in some specific regions, and
multi-occupation is rarely seen in these cases. However, the
Let us consider the simplest case for host systems: a nano- numbers calculated here graphically illustrate the differences
particle (spherical, monodisperse and 10 nm in diameter) and in binding capacities between the nanoparticles and proteins.
a carrier protein (like human serum albumin, with a binding These differences, as we will revise later, have many other
pocket of ≈1.0 nm in diameter), both at a 100 nM concen- implications for modelling and estimating the association of
tration. As a ligand, we will use a well-established guest model molecules with nanoparticulate surfaces.
molecule, methylene blue (MB), that has a rectangle with
dimensions of 0.4 × 0.793 × 1.634 nm.53
First, we calculate the total surface area of a single nano-
Example 2. Effect of component
particle by using the equation that describes the area of a
perfect sphere (A = 4πr2), while for the protein, as an approxi- concentrations on the nanoparticle
mation, we will assume that the pocket has a spherical shape number
as well, thus:
For this example, we will consider a monodisperse sample of
silver nanorods with dimensions of 100 (length) × 25 nm (dia-
Surface area (nm2) meter), like the one shown in Fig. 4. Then, we will use a total
Nanoparticle 314 silver concentration of 0.2 × 10−3 M. AgNP concentration and
Protein 3.14 surface density/concentration can be calculated in a fashion
similar to that described for other nanomaterials like gold and
While for MB, binding can occur through, at least, two silver.67,68 In this calculation, we will assume either complete
different orientations, and consequently two different surface or fifty percent Ag+ reduction to form the nanorods. A valid
areas: approximation for calculating the number of atoms per nano-
particle can be achieved considering that the volume of the calculate how many times the AuNP colloids were concentrated
nanoparticle (VNP) is equal to N times the volume of each atom (named as f ):
(VAtom) as follows:
V NP ¼ N V Atom : π h R 2 ¼ N ð4=3Þ π r 3 f ¼ AbsSPB
A =AbsB
SPB
Using the nanorod diameter, the total number of atoms

contained in a single silver layer in the nanorod can be calcu- where AbsSPB
A is the absorbance of the solution A at the SPB
lated as: wavelength, and AbsSPB
B the absorption of the solution B. Using
this approach to calculate nanogold concentrations, an absorp-
N ¼ 3 ðh R 2 Þ=4 r 3 tion coefficient of ε520 nm = 0.24 ± 0.01 nM−1 cm−1 was calcu-
lated for 12.8 nm AuNP.75
where h is the length (100 nm), R is the nanorod radius
Method 2: Based on the method described by Liu et al.,76
(12.5 nm), and r is the interatomic silver distance, see Fig. 4.
this procedure allows obtaining an estimated value of the
We have used r = 0.234 nm (measured from Fig. 4 high-resolu-
absorption coefficient at 506 nm for a spherical nanogold,
tion TEM images), which gives a total number of ≈920 000
which is independent of the surface composition as:
silver atoms contained in a single nanorod.
½AgNP ¼ ½Ag=N
ε ¼ e ðk ln DþaÞ
The nanorod concentration, [AgNP], can then be calculated
using the equation above, where [Ag] corresponds to the total where ε (M−1 cm−1) is the absorption coefficient, D is the
silver concentration that has been reduced from Ag+ to Ag0. As average nanogold diameter from TEM images, and k and a are
the total silver concentration was equal to 0.2 × 10−3 mol L−1, constants with values of 3.3 ± 0.1 and 10.8 ± 0.3, respectively.76
it renders concentrations of 2.2 and 1.1 × 10−10 M of nanorods Assumptions involved in these calculations are similar to the
for the 100 and 50% reduction of total Ag, respectively. ones mentioned above. Note that the associated errors in the ε
calculation are large (≥%30) if the standard deviation for the
average NP sizes determined by TEM exceeds 5%.
Method 3: This approximation involves a theoretical calcu-
lation according to the work by Lanterna et al.,77 of the corres-
Example 3. Nanoparticle ponding cross-section (CSNP, nm2 per nanoparticle) for a
concentration using different single nanoparticle at the SPB peak in water. The Mie theory
experimental approaches and the dielectric function tabulated by Palik for gold are
employed in this calculation78 and the NP concentration can
Method 1: The concentration of nanoparticles (CNP) can be be then estimated using the Lambert–Beer equation [CNP =
estimated from TEM experiments as described below and also AbsSPB/(CSNPb)], where b is the cell path length in nm. Thus,
by Lewis et al.,72 using the following equation: the CSNP is determined using the following equation:
½AuNP ¼ NNP =NA
¼ Natoms =ðN NA Þ CSNP ¼ Q π ðD=2Þ2
¼ moles Auþ3 l NA =ðN NA Þ
where Q is a theoretical parameter determined, for gold, as
where N is the number of gold atoms per nanoparticle, NNP
(0.07126 × D − 0.2206), where D is the nanoparticle diameter.77
and Natoms are the nanoparticles and gold atoms obtained in a
This method also requires an excellent RSD for the average
liter (l), respectively, and NA is the Avogadro number. N can be
size of the nanomaterial.
calculated from:
Method 4: According to the work by Navarro et al.,79 the
N ¼ ðRNP =r A Þ3 attenuation coefficient (equivalent to the old term ‘extinc-
tion’)80 at the corresponding SPB can be estimated using a
where RNP is the radius of the cluster and rA is the atomic power law (also Mie theory based) as follows:
radius (0.144 nm for fcc Au).73 Assumptions made for this
method are that the volume of the cluster (VNP) is N times the γ
εAuNP
max ¼ A D
volume of the individual atom (VA) (valid for large ‘spherical’
clusters),74 the nanomaterial is highly monodisperse and that
in the synthesis all Au3+ ions are reduced to Au0 atoms. Once D ≤ 85 nm, A = 4.7×104 M−1 cm−1, γ = 3.30 or D > 85 nm, A =
AuNP colloids are concentrated by centrifuging and re-sus- 1.6 × 108 M−1 cm−1, γ = 1.47.
pended in water, the concentration of this NP solution (CNP) Then, the CNP is estimated using the Lambert–Beer
can be estimated according to previous work.67 Briefly, from equation. Although this approach is useful, the associated
the absorption spectra of the concentrated AuNP solution (A) error in the ε calculation is elevated (≥30%) for relative stan-
and the pre-centrifuged AuNP (B) the following ratio is used to dard deviations ≥10% in the NP size as determined by TEM.
Example 4. Effect of nanoparticle size Thus, if one were calculating the association constant for
distribution the binding of MB to nanosilver, using solutions with different
size distributions (centred around 10 nm), the observed discre-
To begin, let us consider the following for studying the effect pancies within the obtained numbers for saturation concen-
of nanoparticle size distribution; in all cases, we will examine trations in the dye can be expected. These differences will
spherical nanosilver. become particularly marked for samples that contain a large
i. 100% of particles monodisperse, with a 10 nm proportion of smaller nanoparticles.
diameter.
ii. 50% of particles with a 10 nm diameter and the rest with
100 nm. Acknowledgements
iii. 50% of particles with a 10 nm diameter and the rest
with 5.0 nm. This work was made possible by funding from NSERC Canada
iv. 50% of particles with a 10 nm diameter, 25% 100 nm to EIA (RGPIN-2015-0632). NLP would like to acknowledge
and 25% 5.0 nm. funding support from CONICET, ANPCyT-PICT 2013-1989, and
v. 50% of particles with a 10 nm diameter, 12.5% 2.5 nm, SECYT-UNC. NLP is a research member of the Consejo
12.5% 50 nm, 12.5% 50 nm, and 12.5% 100 nm. Nacional de Investigaciones Científicas y Técnicas (CONICET)
(i) Is the case typically reported in the literature, where of Argentina.
approximations are done using the average size for a given
nanoparticle population (mean diameter/size). Thus, as we
know the total silver atoms in solution are 0.2 × 10−3 mol L−1, References
the total number of atoms per liter can be calculated using
Avogadro’s number, 6.022 × 1023 atoms per mol, which 1 X. Zhang and K. N. Houk, Acc. Chem. Res., 2005, 38, 379–
renders 1.2044 × 1020 atoms per L. The following number of 385.
atoms for each case, i–v, can be calculated: 2 A. J. T. Smith, X. Zhang, A. G. Leach and K. N. Houk,
J. Med. Chem., 2008, 52, 225–233.
3 K. N. Houk, A. G. Leach, S. P. Kim and X. Zhang, Angew.
Surface area nm2/1015 Chem., Int. Ed., 2003, 42, 4872–4897.
2.5 5.0 10 50 100 Total 4 G. M. Blackburn and A. Garçon, in Biotech. Set, Wiley-VCH
Verlag GmbH, 2008, pp. 401–490.
(i) — — 2.08 — — 2.08 5 S. Mclaughlin, J. Podrebarac, M. Ruel, E. J. Suuronen,
(ii) — — 1.04 — 0.00105 ≈1.04
(iii) — 8.36 1.04 — — 9.40 B. McNeill and E. I. Alarcon, Front. Mater., 2016, 3, DOI:
(iv) — 4.18 1.04 — 0.0005 ≈5.22 10.3389/fmats.2016.00027.
(v) 16.7 2.09 1.04 0.00209 0.000209 ≈20 6 B. K. Nanjwade, H. M. Bechra, G. K. Derkar, F. V. Manvi
and V. K. Nanjwade, Eur. J. Pharm. Sci., 2009, 38, 185–196.
Thus, when comparing the differences between the 7 K. J. Morrow Jr., R. Bawa and C. Wei, Med. Clin. North Am.,
groups, one can see how adding new nanoparticle populations 2007, 91, 805–843.
to the model solution of 10 nm diameter, does affect the total 8 C. L. Ventola, P T, 2012, 37, 582–591.
available nanoparticle surface. Thus, for example, as a general 9 M. L. Etheridge, S. A. Campbell, A. G. Erdman, C. L. Haynes,
trend, adding smaller nanoparticles does increase the S. M. Wolf and J. McCullough, Nanomedicine, 2013, 9, 1–14.
surface area to a greater extent than when larger nano- 10 S. K. Arya, B. Lim and A. R. Rahman, Lab Chip, 2013, 13,
structures are incorporated. This is a direct consequence of 1995–2027.
the fact that smaller nanoparticles have a larger surface area, 11 L. Qin, Y. Xu, H. Han, M. Liu, K. Chen, S. Wang, J. Wang,
relative to the number of atoms per particle, than larger J. Xu, L. Li and X. Guo, J. Colloid Interface Sci., 2015, 460,
nanostructures. 221–229.
Using methylene blue (MB) as a model again, see example 12 J. Hou, D. M. Szaflarski and J. D. Simon, J. Phys. Chem. B,
1, and assuming that the surface area for the dye has an 2013, 117, 4587–4593.
average area of ≈0.8 nm2 (average of orientation a and b), the 13 S. Roy and T. K. Das, J. Appl. Spectrosc., 2015, 82, 598–606.
concentrations of MB required to saturate the surface of the 14 J. S. Gebauer, M. Malissek, S. Simon, S. K. Knauer,
different nanoparticle solutions are: M. Maskos, R. H. Stauber, W. Peukert and L. Treuel,
Langmuir, 2012, 28, 9673–9679.
Methylene blue/[nM] 15 S. Jiang, H.-Z. Liu, W.-L. Cai, A.-m. Bai, Y. Ouyang and
Y.-J. Hu, Luminescence, 2016, 31, 1146–1151.
(i) 4.3 16 S. Pramanik, P. Banerjee, A. Sarkar and S. C. Bhattacharya,
(ii) 2.2
(iii) 20 J. Lumin., 2008, 128, 1969–1974.
(iv) 10 17 M. Iosin, F. Toderas, P. L. Baldeck and S. Astilean, J. Mol.
(v) 42 Struct., 2009, 924–926, 196–200.
18 S. Dominguez-Medina, S. McDonough, P. Swanglap, 44 A. C. Tedesco, D. M. Oliveira, Z. G. M. Lacava,

C. F. Landes and S. Link, Langmuir, 2012, 28, 9131– R. B. Azevedo, E. C. D. Lima, C. Gansau, N. Buske and
9139. P. C. Morais, J. Appl. Phys., 2003, 93, 6704–6706.
19 S. H. Brewer, W. R. Glomm, M. C. Johnson, M. K. Knag and 45 A. C. Tedesco, D. M. Oliveira, Z. G. M. Lacava,
S. Franzen, Langmuir, 2005, 21, 9303–9307. R. B. Azevedo, E. C. D. Lima and P. C. Morais, J. Magn.
20 J. Mariam, S. Sivakami and P. M. Dongre, J. Biomol. Struct. Magn. Mater., 2004, 272–276(Part 3), 2404–2405.
Dyn., 2016, 1–12. 46 P. P. Macaroff, D. M. Oliveira, Z. G. M. Lacava,
21 S. Goy-López, J. Juárez, M. Alatorre-Meda, E. Casals, R. B. Azevedo, E. C. D. Lima, P. C. Morais and
V. F. Puntes, P. Taboada and V. Mosquera, Langmuir, 2012, A. C. Tedesco, IEEE Trans. Magn., 2004, 40, 3027–
28, 9113–9126. 3029.
22 K. G. Thomas, J. Zajicek and P. V. Kamat, Langmuir, 2002, 47 B.-F. Pan, F. Gao and L.-M. Ao, J. Magn. Magn. Mater., 2005,
18, 3722–3727. 293, 252–258.
23 I. Blakey, T. L. Schiller, Z. Merican and P. M. Fredericks, 48 N. Nagarajan, G. Paramaguru, G. Vanitha and
Langmuir, 2010, 26, 692–701. R. Renganathan, J. Chem., 2013, 2013, 7.
24 D. Sotnikov, A. Zherdev and B. Dzantiev, Int. J. Mol. Sci., 49 A. Kathiravan, G. Paramaguru and R. Renganathan, J. Mol.
2015, 16, 907–923. Struct., 2009, 934, 129–137.
25 M. M. Y. Chen and A. Katz, Langmuir, 2002, 18, 2413–2420. 50 A. Bhogale, N. Patel, P. Sarpotdar, J. Mariam, P. M. Dongre,
26 S. Chakraborty, P. Joshi, V. Shanker, Z. A. Ansari, A. Miotello and D. C. Kothari, Colloids Surf., B, 2013, 102,
S. P. Singh and P. Chakrabarti, Langmuir, 2011, 27, 7722– 257–264.
7731. 51 B. Hemmateenejad and S. Yousefinejad, J. Mol. Struct.,
27 G. Wang, X. Liu, C. Yan, G. Bai and Y. Lu, Colloids Surf., B, 2013, 1037, 317–322.
2015, 135, 261–266. 52 E. Alarcon, A. Aspee, E. B. Abuin and E. A. Lissi,
28 A. Dawson and P. V. Kamat, J. Phys. Chem. B, 2000, 104, J. Photochem. Photobiol., B, 2012, 106, 1–17.
11842–11846. 53 X. Zhao, X. Bu, T. Wu, S.-T. Zheng, L. Wang and P. Feng,
29 S. V. Jenkins, A. Srivatsan, K. Y. Reynolds, F. Gao, Y. Zhang, Nat. Commun., 2013, 4, 2344.
C. D. Heyes, R. K. Pandey and J. Chen, J. Colloid Interface 54 X.-R. Song, N. Goswami, H.-H. Yang and J. Xie, Analyst,
Sci., 2016, 461, 225–231. 2016, 141, 3126–3140.
30 L. Shang, S. Brandholt, F. Stockmar, V. Trouillet, M. Bruns 55 N. Goswami, K. Zheng and J. Xie, Nanoscale, 2014, 6,
and G. U. Nienhaus, Small, 2012, 8, 661–665. 13328–13347.
31 C.-S. Tsai, T.-B. Yu and C.-T. Chen, Chem. Commun., 2005, 56 Z. Luo, K. Zheng and J. Xie, Chem. Commun., 2014, 50,
4273–4275. 5143–5155.
32 J. A. Yang, B. J. Johnson, S. Wu, W. S. Woods, J. M. George 57 J. Xie, Y. Zheng and J. Y. Ying, J. Am. Chem. Soc., 2009, 131,
and C. J. Murphy, Langmuir, 2013, 29, 4603–4615. 888–889.
33 K. Chen, Y. Xu, S. Rana, O. R. Miranda, P. L. Dubin, 58 K. Chaudhari, P. L. Xavier and T. Pradeep, ACS Nano, 2011,
V. M. Rotello, L. Sun and X. Guo, Biomacromolecules, 2011, 5, 8816–8827.
12, 2552–2561. 59 A. Baksi, P. L. Xavier, K. Chaudhari, N. Goswami, S. K. Pal
34 G. Y. Tonga, T. Mizuhara, K. Saha, Z. Jiang, S. Hou, R. Das and T. Pradeep, Nanoscale, 2013, 5, 2009–2016.
and V. M. Rotello, Tetrahedron Lett., 2015, 56, 3653–3657. 60 J. Zheng, J. T. Petty and R. M. Dickson, J. Am. Chem. Soc.,
35 C. C. Fleischer and C. K. Payne, J. Phys. Chem. B, 2014, 118, 2003, 125, 7780–7781.
14017–14026. 61 B. Santiago González, M. J. Rodríguez, C. Blanco, J. Rivas,
36 S. R. Kavitha, M. Umadevi, P. Vanelle, T. Terme and M. A. López-Quintela and J. M. G. Martinho, Nano Lett.,
O. Khoumeri, Eur. Phys. J., 2014, 68, 308. 2010, 10, 4217–4221.
37 R. Revathi, A. Rameshkumar and T. Sivasudha, 62 H. Kawasaki, K. Hamaguchi, I. Osaka and R. Arakawa, Adv.
Spectrochim. Acta, Part A, 2016, 152, 192–198. Funct. Mater., 2011, 21, 3508–3515.
38 R. Choudhury, A. Purkayastha, D. Debnath and T. K. Misra, 63 S. McLaughlin, M. Ahumada, W. Franco, T.-F. Mah,
J. Mol. Recognit., 2016, 29, 452–461. R. Seymour, E. J. Suuronen and E. I. Alarcon, Nanoscale,
39 Y. Tauran, A. Brioude, B. Kim, F. Perret and A. Coleman, 2016, 8, 19200–19203.
Molecules, 2013, 18, 5993–6007. 64 H. Poblete, A. Agarwal, S. S. Thomas, C. Bohne,
40 S. Zhang, H. Yang, X. Ji and Q. Wang, Spectrochim. Acta, R. Ravichandran, J. Phospase, J. Comer and E. I. Alarcon,
Part A, 2016, 154, 33–41. Langmuir, 2016, 32, 265–273.
41 A. Bhogale, N. Patel, J. Mariam, P. M. Dongre, A. Miotello 65 R. J. Hunter, in Zeta Potential in Colloid Science, Academic
and D. C. Kothari, Colloids Surf., B, 2014, 113, 276–284. Press, 1981, pp. v–vi.
42 T. Soukka, H. Härmä, J. Paukkunen and T. Lövgren, Anal. 66 M. Rahman, S. Laurent, N. Tawil, L. H. Yahia and
Chem., 2001, 73, 2254–2260. M. Mahmoudi, in Protein-Nanoparticle Interactions: The Bio-
43 P. Canoa, R. Simón-Vázquez, J. Popplewell and Á. González- Nano Interface, Springer, Berlin, Heidelberg, 2013, pp.
Fernández, Biosens. Bioelectron., 2015, 74, 376–383. 21–44.
67 N. L. Pacioni, M. González-Béjar, E. I. Alarcón, 80 IUPAC Compendium of Chemical Terminology http://gold-

K. L. McGilvray and J. C. Scaiano, J. Am. Chem. Soc., 2010, book.iupac.org/A00516.html.
132, 6298–6299. 81 P. Thordarson, Chem. Soc. Rev., 2011, 40, 1305–
68 H. de Alwis Weerasekera, M. Griffith and E. I. Alarcon, in 1323.
Silver Nanoparticle Applications: In the Fabrication and 82 K. A. Connors, Binding Constants: The Measurement of
Design of Medical and Biosensing Devices, ed. E. I. Alarcon Molecular Complex Stability, Wiley, 1987.
and K. Udekwu and M. Griffith, Springer, UK, 2015, pp. 83 E. I. Alarcon, C. J. Bueno-Alejo, C. W. Noel,
93–125. K. G. Stamplecoskie, N. L. Pacioni, H. Poblete and
69 A. S. Mahadevi and G. N. Sastry, Chem. Rev., 2016, 116, J. C. Scaiano, J. Nanopart. Res., 2013, 15, 1374.
2775–2825. 84 L. Shang, S. Brandholt, F. Stockmar, V. Trouillet, M. Bruns
70 E. I. Alarcon, B. Vulesevic, A. Argawal, A. Ross, P. Bejjani, and G. U. Nienhaus, Small, 2014, 10, 1667–1667.
J. Podrebarac, R. Ravichandran, J. Phopase, E. J. Suuronen 85 B. Valeur and M. N. Berberan-Santos, in Molecular
and M. Griffith, Nanoscale, 2016, 8, 6484–6489. Fluorescence, Wiley-VCH Verlag GmbH & Co. KGaA, 2012,
71 V. V. Hoang and D. Ganguli, Phys. Rep., 2012, 518, 81–140. pp. 263–283.
72 D. Lewis, T. Day, J. Macpherson and Z. Pikramenou, Chem. 86 F. Loosli, L. Vitorazi, J.-F. Berret and S. Stoll, Environ. Sci.:
Commun., 2006, 1433–1435. Nano, 2015, 2, 541–550.
73 K. L. online, Tables of Physical & Chemical Constants, 87 A. Gourishankar, S. Shukla, K. N. Ganesh and M. Sastry,
http://www.kayelaby.npl.co.uk. J. Am. Chem. Soc., 2004, 126, 13186–13187.
74 G. L. Hornyak, J. Dutta, H. F. Tibbals and A. K. Rao, 88 H. Joshi, P. S. Shirude, V. Bansal, K. N. Ganesh and
Introduction to Nanoscience, CRC Press, Boca Raton, FL, M. Sastry, J. Phys. Chem. B, 2004, 108, 11535–11540.
2008. 89 L. Qin, Y. Xu, H. Han, M. Liu, K. Chen, S. Wang, J. Wang,
75 N. L. Pacioni and A. V. Veglia, Sens. Actuators, B, 2016, 228, J. Xu, L. Li and X. Guo, J. Colloid Interface Sci., 2015, 460,
557–564. 221–229.
76 X. Liu, M. Atwater, J. Wang and Q. Huo, Colloids Surf., B, 90 K. Chen, Y. Xu, S. Rana, O. R. Miranda, P. L. Dubin,
2007, 58, 3–7. V. M. Rotello, L. Sun and X. Guo, Biomacromolecules, 2011,
77 A. E. Lanterna, E. A. Coronado and A. M. Granados, J. Phys. 12, 2552–2561.
Chem. C, 2012, 116, 6520–6529. 91 N. S. Abadeer, G. Fülöp, S. Chen, M. Käll and C. J. Murphy,
78 E. D. Palik, in Handbook of Optical Constants of Solids, ACS Appl. Mater. Interfaces, 2015, 7, 24915–
Academic Press, Boston, 1998, pp. 3–19. 24925.
79 J. R. G. Navarro and M. H. V. Werts, Analyst, 2013, 138, 92 A. Doménech-Carbó, M. González-Béjar and J. Pérez-Prieto,
583–592. Electroanalysis, 2015, 27, 2302–2312.

Analyst 2017

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Analyst 2017

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Analyst 2017

Uploaded by

Copyright:

Available Formats

Analyst

Association models for binding of molecules to

Manuel Ahumada studied chem- Eduardo Lissi Gervaso obtained

Ana Maria Montagut obtained Francisco H. Valenzuela

Natalia L. Pacioni obtained her Emilio I. Alarcon received his

This journal is © The Royal Society of Chemistry 2017

Analyst, 2017, 142, 2067–2089 | 2071

Gold–3,6,9,12- β-Lactoglobulin BLGB 10 — — — Single set of 210 31 33

2072 | Analyst, 2017, 142, 2067–2089

This journal is © The Royal Society of Chemistry 2017

potential Polydispersity constant Occupation

Silver [32]-Membered octa ammonium 9.62 — — — Benesi–Hildebrand 263 — 38

This journal is © The Royal Society of Chemistry 2017

Analyst, 2017, 142, 2067–2089 | 2073

Fig. 1 Representative association constant distributions for molecules

iii. Underestimation of the nanoparticle concentration, and

of each of the above-mentioned points including some practi-

cal examples for illustrating the consequences of each of these

2.1 Metal clusters as binding platforms

Special cases related to the ligand size-dependence are metal

nanoclusters (NCs), which are characterized by having a small

core size (<2 nm), defined molecular structure, discrete elec-

tronic transitions and strong photoluminescence. The litera-

Human serum albumin (HSA)

Bovine serum albumin (BSA)

together with the use of biomolecules has catapulted the use

nanoparticles, the size of the ligand will modulate the size of

number of atoms was seen in native lactotransferin

core has around 12 Au atoms;59 while for polymers like poly

(amidoamine) (PAMAM; ≈5.0 kDa) and poly(N-vinylpyrroli-

done) (PVP; ≈10 kDa), 8 and 2–3 (main species) Au atoms

were, respectively, found in the metal core.60,61 Furthermore,

the metal core size is highly dependent on the conformation

Scheme 1 Top: Cartoon representation of a multi-step capping agent

eﬀect on the resulting association constant for a given mole-

UV-Vis spectroscopy [NP] = constant; [Ligand] = variable Phenyl isothiocyanate (PITC) 22

3.2 Fluorescence spectroscopy

have some diﬀerences including the formation of organized

Note that these numbers correspond to a maximum

Using the nanorod diameter, the total number of atoms

18 S. Dominguez-Medina, S. McDonough, P. Swanglap, 44 A. C. Tedesco, D. M. Oliveira, Z. G. M. Lacava,

67 N. L. Pacioni, M. González-Béjar, E. I. Alarcón, 80 IUPAC Compendium of Chemical Terminology http://gold-

You might also like