Albumin-based nanoparticles as magnetic resonance contrast agents: I. Concept, first syntheses and...

26
ORIGINAL PAPER Albumin-based nanoparticles as magnetic resonance contrast agents: II. Physicochemical characterisation of purified and standardised nanoparticles A. A. Abdelmoez G. C. Thurner E. A. Wallno ¨fer N. Klammsteiner C. Kremser H. Talasz M. Mrakovcic E. Fro ¨hlich W. Jaschke P. Debbage Accepted: 30 June 2010 / Published online: 14 July 2010 Ó Springer-Verlag 2010 Abstract We are developing a nanoparticulate histo- chemical reagent designed for histochemistry in living animals (molecular imaging), which should finally be useful in clinical imaging applications. The iterative development procedure employed involves conceptual design of the reagent, synthesis and testing of the reagent, then redesign based on data from the testing; each cycle of testing and development generates a new generation of nanoparticles, and this report describes the synthesis and testing of the third generation. The nanoparticles are based on human serum albumin and the imaging modality selected is mag- netic resonance imaging (MRI). Testing the second particle generation with newly introduced techniques revealed the presence of impurities in the final product, therefore we replaced dialysis with diafiltration. We introduced further testing methods including thin layer chromatography, arsenazo III as chromogenic assay for gadolinium, and sev- eral versions of polyacrylamide gel electrophoresis, for physicochemical characterisation of the nanoparticles and intermediate synthesis compounds. The high grade of chemical purity achieved by combined application of these methodologies allowed standardised particle sizes to be achieved (low dispersities), and accurate measurement of critical physicochemical parameters influencing particle size and imaging properties. Regression plots confirmed the high purity and standardisation. The good degree of quantitative physicochemical characterisation aided our understanding of the nanoparticles and allowed a conceptual model of them to be prepared. Toxicological screening demonstrated the extremely low toxicity of the particles. The high magnetic resonance relaxivities and enhanced mechanical stability of the particles make them an excellent platform for the further development of MRI molecular imaging. Keywords Albumin nanoparticles MRI Gadolinium Targeting PEGylation Molecular imaging Introduction Histochemistry in living animals has become known as molecular imaging. Molecular imaging by magnetic reso- nance imaging (MRI) (MRMI) promises early detection of malignancies (Harisinghani et al. 2003; Jaffer and Weiss- leder 2005) and of vulnerable atherosclerotic plaques (Winter et al. 2003), in addition to monitoring of drug distribution (Saito et al. 2004; Griffiths and Glickson 2000) A. A. Abdelmoez and G. C. Thurner contributed equally to this work. A. A. Abdelmoez G. C. Thurner E. A. Wallno ¨fer C. Kremser W. Jaschke Department of Radiology, Innsbruck Medical University, Anichstrasse 35, 6020 Innsbruck, Austria N. Klammsteiner P. Debbage (&) Department of Anatomy, Histology and Embryology, Innsbruck Medical University, Mu ¨llerstrasse 59, 6020 Innsbruck, Austria e-mail: [email protected] H. Talasz Biozentrum of the Medical University Innsbruck, Section for Clinical Biochemistry, Fritz-Pregl-Straße 3, 6020 Innsbruck, Austria M. Mrakovcic E. Fro ¨hlich Center for Medical Research, Stiftingtalstrasse 24, 8010 Graz, Austria Present Address: A. A. Abdelmoez Department of Pharmaceutical Organic Chemistry, Faculty of Pharmacy, Assiut University, Assiut, Egypt 123 Histochem Cell Biol (2010) 134:171–196 DOI 10.1007/s00418-010-0726-6

Transcript of Albumin-based nanoparticles as magnetic resonance contrast agents: I. Concept, first syntheses and...

ORIGINAL PAPER

Albumin-based nanoparticles as magnetic resonance contrastagents: II. Physicochemical characterisation of purifiedand standardised nanoparticles

A. A. Abdelmoez • G. C. Thurner • E. A. Wallnofer •

N. Klammsteiner • C. Kremser • H. Talasz •

M. Mrakovcic • E. Frohlich • W. Jaschke • P. Debbage

Accepted: 30 June 2010 / Published online: 14 July 2010

� Springer-Verlag 2010

Abstract We are developing a nanoparticulate histo-

chemical reagent designed for histochemistry in living

animals (molecular imaging), which should finally be useful

in clinical imaging applications. The iterative development

procedure employed involves conceptual design of the

reagent, synthesis and testing of the reagent, then redesign

based on data from the testing; each cycle of testing and

development generates a new generation of nanoparticles,

and this report describes the synthesis and testing of the

third generation. The nanoparticles are based on human

serum albumin and the imaging modality selected is mag-

netic resonance imaging (MRI). Testing the second particle

generation with newly introduced techniques revealed the

presence of impurities in the final product, therefore we

replaced dialysis with diafiltration. We introduced further

testing methods including thin layer chromatography,

arsenazo III as chromogenic assay for gadolinium, and sev-

eral versions of polyacrylamide gel electrophoresis, for

physicochemical characterisation of the nanoparticles

and intermediate synthesis compounds. The high grade of

chemical purity achieved by combined application of these

methodologies allowed standardised particle sizes to be

achieved (low dispersities), and accurate measurement of

critical physicochemical parameters influencing particle size

and imaging properties. Regression plots confirmed the high

purity and standardisation. The good degree of quantitative

physicochemical characterisation aided our understanding

of the nanoparticles and allowed a conceptual model of them

to be prepared. Toxicological screening demonstrated the

extremely low toxicity of the particles. The high magnetic

resonance relaxivities and enhanced mechanical stability of

the particles make them an excellent platform for the further

development of MRI molecular imaging.

Keywords Albumin nanoparticles � MRI � Gadolinium �Targeting � PEGylation � Molecular imaging

Introduction

Histochemistry in living animals has become known as

molecular imaging. Molecular imaging by magnetic reso-

nance imaging (MRI) (MRMI) promises early detection of

malignancies (Harisinghani et al. 2003; Jaffer and Weiss-

leder 2005) and of vulnerable atherosclerotic plaques

(Winter et al. 2003), in addition to monitoring of drug

distribution (Saito et al. 2004; Griffiths and Glickson 2000)

A. A. Abdelmoez and G. C. Thurner contributed equally to this work.

A. A. Abdelmoez � G. C. Thurner � E. A. Wallnofer �C. Kremser � W. Jaschke

Department of Radiology, Innsbruck Medical University,

Anichstrasse 35, 6020 Innsbruck, Austria

N. Klammsteiner � P. Debbage (&)

Department of Anatomy, Histology and Embryology, Innsbruck

Medical University, Mullerstrasse 59, 6020 Innsbruck, Austria

e-mail: [email protected]

H. Talasz

Biozentrum of the Medical University Innsbruck,

Section for Clinical Biochemistry, Fritz-Pregl-Straße 3,

6020 Innsbruck, Austria

M. Mrakovcic � E. Frohlich

Center for Medical Research, Stiftingtalstrasse 24,

8010 Graz, Austria

Present Address:A. A. Abdelmoez

Department of Pharmaceutical Organic Chemistry,

Faculty of Pharmacy, Assiut University, Assiut, Egypt

123

Histochem Cell Biol (2010) 134:171–196

DOI 10.1007/s00418-010-0726-6

and application as surrogates for clinical testing (Rehman

and Jayson 2005). MRMI is therefore under development

by numerous groups around the world (Flacke et al. 2001;

Hengerer and Grimm 2006; Mulder et al. 2006; Utsumi

et al. 2006). MRMI requires use of nanoparticles as

amplifiers for specific signals (Debbage and Jaschke 2008).

Since nanoparticles are novel agents with largely unknown

toxicological and immunological potential (Debbage 2009;

Stollenwerk et al. 2010; Robbens et al. 2010; Vega-Villa

et al. 2008; Suh et al. 2009), adequate development and

testing in iterative developmental cycles are necessary to

develop safe and effective agents for clinical use. This

paper is the second of a series describing our design,

synthesis and optimisation of successive generations of

albumin-based nanoparticles for eventual application in

clinical imaging and therapy. We recently reported syn-

thesis and characterisation of two generations of nanopar-

ticles formed from HSA emulsified with polylactic acid,

the nanoparticles of approximately 30 nm diameter, and

each bearing several hundred gadolinium-DTPA-chelates

(Stollenwerk et al. 2010). The major advance achieved at

this stage was upscaling the synthesis to produce 3–5 g of

nanoparticles as product. However, the variability of the

nanoparticles was too high to allow precise interbatch

comparisons, and much too high to consider use of the

particles in any clinical application (Stollenwerk et al.

2010). In this paper, we address the question of standar-

dising the nanoparticles, aiming to reduce the intra- and

interbatch variability and to increase the purity of the

batches. We considered it likely that the variability had two

possible causes, the first being the variability in the starting

raw materials, and the other being due to the presence of

educts (starting materials) remaining as contaminating

impurities in the final nanoparticle preparations. To detect

chemical impurities, we extended the range of analytical

techniques applied at different stages of the synthesis

procedure. A further major concern was the stability of the

nanoparticles in the physiological environments within

living animals. We noted earlier that complex structures

such as nanoparticles can show different types of insta-

bility. We focus here on one particular type of stability,

namely mechanical stability. We aimed to increase the

internal cohesion of the nanoparticles, so that they would

retain their integrity as individual particles and not break

up into fragments during passage through the bloodstream

or during intercellular processing. We describe the cross-

linked nanoparticles, and also procedures for testing the

degree of nanoparticle stability. Finally, for protein-based

nanoparticles, it is essential to ensure that the protein

molecules retain their native configuration, in order to

preserve their specific functions [for HSA, this is the

binding and transport of small molecules (Fehske et al.

1981; Frokjaer and Otzen 2005; Goldwasser and Feldman

1997; Gonzalez and Kannewurf 1998; Griffel and Kaufman

1992; Kragh-Hansen 1981; Montero et al. 2007; Peters

1985; Putnam 1984; Rainey and Read 1994)], and to avoid

the potentially dangerous effects of denaturation, such as

triggering amyloid formation (Schnabel 2010; Goldschmidt

et al. 2010; Balbirnie et al. 2001; Nelson et al. 2005).

In the first paper of this series, we described the syn-

thesis and characterisation of ‘‘naked’’ nanoparticles,

which were derivatized in only one way, by incorporation

of gadolinium to allow detection of the particles by MRI.

The ultimate aim of this work is to develop specific

histochemical staining in living animals, which requires

targeting of the nanoparticles to specified sites within

living tissues. In this paper, we explore the result of

adding targeting groups to the particles, aiming to exploit

well-standardised, mechanically stable nanoparticles as a

platform for targeting applications. Adjunct to this, we

check the nanoparticle properties after attachment of

PEG chains (Veronese 2001; Zalipsky 1995a), which are

frequently used to confer ‘‘stealth’’ properties on nano-

particles, reducing nanoparticle uptake into the reticulo-

endothelial system (Abuchowski et al. 1977; Zalipsky

1995b).

This paper describes four further steps in our develop-

ment of nanoparticles for use as intravital histochemical

stains. As conceptual background, the potential application

of the nanoparticles for clinical use is always present, and a

major criterion of success is the possibility of transferring

the synthesis protocols to industrial production. Assess-

ment of this possibility is an integral part of this paper.

Materials and methods

Materials

In addition to materials listed in our previous paper (Stol-

lenwerk et al. 2010), charcoal, sodium chloride, caprylic

acid, gadolinium chloride (GdCl3), 2,7-bis(o-arseno-

phenylazo)-1,8-dihydroxynaphthalene-3,6-disulfonic acid

(arsenazo III), sodium cyano-borohydride (NaCNBH3),

iodine and ninhydrine were purchased from Sigma–Aldrich

(Munich, Germany). Rotilabo� injection filter (PP-Geha-

use, sterile, PVDF 0.2 lm), Rotilabo� paper filter (type

600P, ø 185 mm) were purchased from Carl Roth GmbH &

Co (Karlsruhe, Germany). Materials used for polyacryl-

amide gel electrophoresis (PAGE) (SDS, Native, pI),

namely the gels (Ready Gel� precast gels, 10 well comb,

30 ll load volume), the standards, the running and the

sample buffers were purchased from BioRad (Vienna,

Austria). The sheets for thin layer chromatography (TLC)

and the membranes for diafiltration were purchased from

VWR (Vienna, Austria). Adjustments of pH values were

172 Histochem Cell Biol (2010) 134:171–196

123

done with a pH 211 Microprocessor pH meter, Hanna

instruments (Kehl am Rhein, Germany); the electrode

cleaning solution HI 7073, the electrode storage solution

HI 70300 and the pH meter calibration buffers (pH 4, pH 7,

pH 10) were purchased from Carl Roth GmbH & Co

(Karlsruhe, Germany). Throughout this work, the water

used was purified Millipore water (Millipore, Billerica,

MA, USA).

Purification and extraction procedures

To purify the reaction products diafiltration was intro-

duced, replacing simple dialysis (Stollenwerk et al. 2010).

Diafiltration was carried out after each derivatization step

to separate the reaction product (conjugate, nanoparticles)

from contaminating component molecules (HSA-DTPA-

Gd and also small molecular weight species), by use of a

Minimate TFF System, and Minimate TFF membrane

cassettes with membranes of Omega� type (Pall Life Sci-

ences, MI, USA), having cutoffs of 10, 100 and 500 kDa.

To ensure good purification distilled water was added to an

amount at least seven times larger than that used as starting

volume of the nanoparticle solution, additionally pressure

was applied in the range 10–20 psi and the sample solution

was stirred. The quality and end point of diafiltration were

checked by TLC and purification was stopped as soon as no

traces of contaminants remained. If not stated differently

the procedure of diafiltration always was carried out

according to this protocol.

Preparation of the HSA-DTPA-Gd conjugates

Cleaning and stabilisation of HSA

Albumin was prepared by the method of Chen (1967) and

then stabilised by addition of octanoate according to

Shrake et al. (2005): a solution containing 50 mg/ml of

human serum albumin in distilled water was prepared at

room temperature. Then charcoal powder was added under

continuous stirring (HSA:charcoal = 2:1 w/w) and the pH

was adjusted to 3 with 1 M HCl. The suspension was

placed in the fridge at 4�C and stirred magnetically for 1 h.

Afterwards the suspension was filtered through a folded-

filter paper and centrifuged for 10 min at 5,000g. The

supernatant was further filtered through a 0.2 lm filter. The

pH of the solution containing the cleaned HSA molecules

was raised to 4 with 0.1 M NaOH. NaCl, predissolved in

distilled water, was added to a final concentration of

150 mM and the HSA concentration was adjusted to

30 mg/ml. Then Na-Caprylate was added (30 mM Na-

Caprylate per ml nanoparticle sample) and the pH was

slowly increased to 10 by addition of 3 M NaOH. After

stirring magnetically for 1 h at room temperature the pH

was decreased to 7 by addition of 1 M HCl and stirred

overnight at room temperature.

Conjugation of DTPABA to HSA

The procedure described earlier (Stollenwerk et al. 2010)

was slightly modified, omitting the suspension of DTPABA

in DMSO. The following procedure was used instead: the

pH of the protein solution (30 mg/ml as measured by UV

spectrometry) was adjusted to 8.5 by means of 3 M NaOH

and a 50-fold molar excess of solid DTPABA was added

portionwise. The pH of 8.5 was maintained throughout the

chelate addition by adequate provision of 3 M NaOH. After

the reaction mixture was stirred for 2 h at RT, diafiltration

was performed using a membrane of 100 kDa cutoff. The

resulting HSA-DTPA conjugates were now ready for che-

lation of a metal, in this investigation gadolinium (Gd).

Preparation of trisodium bis(nitrilotriacetate) gadolinate

(Na3Gd(NTA)2)

In our previous work (Stollenwerk et al. 2010), gadolinium

oxide was used as starting material. In the work reported

here, this was replaced with gadolinium chloride, as fol-

lows: 4.6 g of gadolinium chloride hexahydrate

(1.24e-02M) were dissolved in 60 ml water and 6.38 g

Na3NTA (2.48e-02M) were added under continuous stir-

ring. After complete dissolution of Na3NTA, the pH was

lowered from 7.5 to 6.0 using 1 M HCl, the volume was

then made up to 100 ml with water and the solution was

stored at 4�C.

GdCl3 þ 2Na3NTA! Gd NTAð Þ2Na3 þ 3NaCl

The product was assayed for the presence of non-

chelated gadolinium using arsenazo III.

Gadolinium complexation to HSA-DTPA

The principle of this chelation is shown in our earlier paper

(Stollenwerk et al. 2010); during the work reported here the

chelation of the conjugate was performed as follows: A 20-

or 50-fold molar excess of Na3[Gd(NTA)2] solution

(116 mM, pH 6.0) was added to HSA-DTPA in 0.1 M

citrate buffer. The sample was then stirred for 24 h at 4�C

followed by diafiltration using a membrane of 100 kDa

cutoff. A subsequent arsenazo III assay was carried out to

check there was no remaining free gadolinium.

Additionally, the following assays were carried out to

examine any uptake of gadolinium by HSA which was

independent of the chelator:

1. Gd-NTA was added to albumin bearing no DTPA, and

the resulting mixture was passed through 14 cycles

of diafiltration. Afterwards the Gd:HSA ratio was

Histochem Cell Biol (2010) 134:171–196 173

123

quantified by atomic absorption spectrometry (AAS)

and Pierce assays.

2. Gd-NTA was added to albumin after prior incubation of

the albumin with 1 M citrate buffer (HSA:cit-

rate = 10:1 v/v). The resulting mixture was passed

through 14 cycles of diafiltration. Afterwards the

Gd:HSA ratio was quantified by AAS and Pierce assays.

Preparation of the nanoparticles

Emulsification

A mixture of PLA plus HSA-DTPA-Gd, based on 1:10 M

ratio PLA:HSA, was emulsified, using the protocol

described in detail for series II by Stollenwerk et al. (2010).

The resulting turbid suspension was filtered through paper

filters to clear the solution prior to diafiltration using a

membrane of 500 kDa cutoff.

Crosslinking

The nanoparticles were crosslinked by addition of 1%

glutaraldehyde solution (glutaraldehyde:nanoparticle ratio

1:10 v/v) and stirred overnight at room temperature followed

by diafiltration using a membrane with 500 kDa cutoff.

Additionally, the following variant procedures were

carried out to examine the nanoparticle aggregation

behaviour and stability:

1. Different concentrations of glutaraldehyde (2–5%)

were added (glutaraldehyde:nanoparticle ratio 1:10 v/

v) and stirred at RT for different durations (1–12 h).

(a) The surplus glutaraldehyde in the crosslinked

nanoparticle samples was quenched with 1% of

aqueous NH4Cl solution (NH4Cl:nanoparticle

ratio was 1:5 v/v). After further stirring for 1 h

the nanoparticle solutions were diafiltered using a

membrane with 500 kDa cutoff.

(b) After crosslinking NaCNBH3 was added to stabi-

lise the Schiff bases formed by glutaraldehyde,

reducing them to covalency according to the

methods described by Means and Feeney (1995).

Briefly, for 10 mM glutaraldehyde 12.5 mM NaC-

NBH3 were added and stirred for 2 h at room

temperature. Then the sample was diafiltered using

a membrane with 500 kDa cutoff.

Physicochemical and immunohistochemical

characterisation: analytical techniques

Nanoparticles were characterised by use of several techniques

described in detail in Stollenwerk et al. (2010), as follows:

1. The molecular weights of HSA and of HSA-DTPA-Gd

conjugate (referred to in the following simply as

‘‘conjugate’’) were estimated by SDS–PAGE.

2. The Gd content of conjugates and nanoparticles was

determined by AAS.

3. The protein content of conjugates and nanoparticles

was determined by the Pierce reaction, and by UV

spectrometry.

4. MR relaxivity properties of conjugates and nanopar-

ticles were measured by use of an inversion recovery

sequence for T1, and by use of a CPMG-type multi

echo spin-echo sequence for T2.

5. Electrical charge properties of the nanoparticles were

determined by measuring their electrophoretic mobi-

lity using a PSS NICOMPTM 380 DLS/ZLS.

6. The presence of albumin in its natural configuration

was tested by checking the presence of epitopes using

immunohistochemistry to detect HSA, carried out on

formvar-coated grids.

7. Size measurements of the nanoparticles were made

from negatively contrasted preparations on formvar-

coated grids by transmission electron microscopy

(TEM), and analysed by use of the Metamorph

programme (Zeiss, Germany). Ultrastructural analyses

were also carried out on pelleted nanoparticles fixed in

glutaraldehyde, postfixed in osmium tetroxide, embed-

ded in Epon and ultrathin-sectioned.

8. Size measurements were also carried out by photon

correlation spectroscopy (PCS); the PCS data was

collated with the TEM data.

9. Nanoparticle masses were calculated by use of the

following formulae, in which diameter d (nm) of the

particles; conjugate density of HSA-DTPA-Gd =

1.4 g/cm3; packing density d of albumin molecules in

the nanoparticles *0.40. The results were obtained

first as nanoparticle mass Mg measured in grams, then

converted to nanoparticle mass MDa expressed as

Daltons:

Mg ¼ 4=3p d=2ð Þ3�10�21 � 1:4� dh i

g;

MDa ¼ Mg=1:66054e�24� �

Da:

10. Nanoparticle structural stability was tested by

exposing them to ultrasound, to detergents, to heat,

and to various combinations of these agents.

11. Nanoparticle cytotoxicity was examined by assess-

ing the effect of incubating the endothelial cell line

EAhy926 with nanoparticles, using the activity of

intracellular dehydrogenases, intracellular ATP con-

tent, and lactate dehydrogenase release as indicators

for cellular damage; these tests are in conformity

with the ‘‘Biological Evaluation of Medical Devices-

part 5: tests for in vitro cytotoxicity guidelines (ISO

174 Histochem Cell Biol (2010) 134:171–196

123

10993-5:1999)’’. Blood smears were also evaluated

to identify changes in platelets due to treatment with

nanoparticles. Whole blood sampled in citrate tubes

was incubated, within 10 min after venipuncture, for

10 min at room temperature with an equal volume of

particles diluted in phosphate-buffered saline (PBS).

Whole blood incubated with ADP (20 lM), for

2 min at room temperature, served as positive

control. After incubation, blood smears were pre-

pared and fixed for 5 min in 100% methanol.

Subsequently, smears were stained, firstly for

3 min with May–Grunwald (Gatt-Koller, Austria)

staining solution, secondly for 1–2 min with May–

Grunwald solution diluted 1:2 in PBS and thirdly for

40 min with the Giemsa staining reagent diluted

1:10 in distilled water. After rinses in distilled water

the slides were coverslipped and viewed under an

Olympus IX51 microscope.

12. Nanoparticle hemocompatibility was examined by

assessing the effect of incubating human erythrocyte

suspensions with the nanoparticles and determining

the degree of hemolysis. Plasma from human volun-

teers was also incubated with the nanoparticles, and

levels of complement C3a and of prothrombin and D-

dimer measured as indicators of activation of the

complement system and of coagulation; these tests

are in conformity with the ‘‘Biological Evaluation of

Medical Devices-part 4: selection of tests for inter-

action with blood (ISO 10993-4:2002)’’, and were

approved by the local Ethics Committee.

13. The yields of nanoparticles were determined for each

batch, to allow estimation of interbatch variability

and of overall synthesis efficiency.

The above analytical procedures were carried out

exactly as described in our previous paper (Stollenwerk

et al. 2010), to allow precise comparison with the nano-

particles of series I and series II described in that report.

The nanoparticles described in the present paper can

therefore be considered as series III in our ongoing trans-

lational research in iterative design, synthesis and optimi-

sation of successive generations of albumin-based

nanoparticles. In order to gain closer control over our

synthetic procedures, however, we introduced a suite of

further analytical techniques aimed at real-time analysis of

intermediate-product and product quality during the syn-

thetic procedures. In this way it was possible to vary sen-

sitive parameters rapidly in a directed fashion and thus to

fine-tune batch optimisation. These newly introduced

assays are described in detail below; furthermore, the

assays described above were also sometimes applied in

modified form, which will be noted as appropriate in the

subsequent text. The newly introduced assays were as

follows:

14. Preparation of lectin-targeted nanoparticles Lyco-

persicon esculentum agglutinin (LEA) was extracted

and purified from tomato fruits as described in

our earlier paper (Paschkunova-Martic et al. 2005).

A nanoparticle colloidal solution in PBS (pH

8.0, 0.15 M) was used. N-(3-dimethylaminopropyl)-

N0-ethyl-carbodiimide hydrochloride was added

(0.63 mg carbodiimide per mg nanoparticle) in order

to activate the carboxylic groups on the albumin

molecules of the nanoparticle surface. The mixture

was stirred for 2 h at room temperature and, in some

batches, excess carbodiimide reagent was removed,

then LEA presuspended in distilled water was added

dropwise to the activated nanoparticles (LEA:nano-

particle ratio 1:20 w/w). The pH was adjusted to 6

with 1 M HCl and the sample was stirred over night

at room temperature, thus coupling the LEA to the

nanoparticles. Aliquots of the resulting nanoparticle

colloidal solutions were analysed by PCS, TEM, and

by SDS gel electrophoresis.

15. Preparation of PEGylated nanoparticles a-Methoxy-

x-carboxylic acid succinimidyl ester polyethylene

glycols having molecular weights of 5, 10 and

20 kDa, respectively, were used (Iris Biotech GmbH

Marktredwitz, Germany). The activated PEGs were

added in solid form to the nanoparticle samples

(HSA:PEG molar ratio 1:1) and stirred for 30 min at

room temperature. Afterwards the nanoparticles were

crosslinked with 1% glutaraldehyde (1:10 v/v) and

stirred over night at room temperature.

16. TLC was used to determine the purity of the starting

materials, conjugates, and nanoparticles. TLC was

carried out on Silica Gel 60 UV254 sheets (Polygram�

Sil G/UV254, 0.2 mm; Macherey–Nagel, Germany).

The solvent system employed was citric acid and

sodium citrate in a 1:2 v/v ratio. The samples were

spotted onto the sheets using micropipettes. After the

spots were completely dry the sheets were developed

in the citrate solvent system. Afterwards the sheets

were dried and detection of the samples was

performed using UV light at 254 and 366 nm as well

as iodine vapour or ninhydrine spraying-reagent.

Some sheets were prepared for detection of free

gadolinium, by brief immersion in the Arsenazo III

solution (see below).

17. Colorimetric testing with Arsenazo III (10 lmol

solution in 0.1 M acetate buffer, pH 5) was per-

formed according to a protocol from Nagaraja et al.

(2006). Briefly, 200 ll of arsenazo III were spiked

with 50 ll of conjugate or nanoparticle samples; a

Histochem Cell Biol (2010) 134:171–196 175

123

change in the colour of arsenazo III then indicated the

chemical state of gadolinium (chelated metal gave a

purple colour, and non-chelated metal gave a green

colour).

18. Pyknometry was employed to measure the density of

conjugates and of nanoparticles. Conjugates as well

as nanoparticles were measured as liquids and their

densities were calculated based on their concentra-

tions. Concentrations of HSA, HSA-DTPA-Gd and of

nanoparticles were measured by UV spectrometry at

280 nm. Density measurements were carried out

using a pyknometer containing a thermometer (Brand

GmbH und Co KG, 97861 Wertheim/Main, Ger-

many; nominal volume 10.0 cm3, certified measured

volume of 9.8162 cm3). The density was calculated

as mass/volume at 20�C.

19. Polyacrylamide gel electrophoresis (PAGE) for all

PAGE methods described here the Mini-PROTEAN�

Tetra Cell from BioRad (Vienna, Austria) as well as

their gels, standards, running and sample buffers were

used. Only the fixative, staining and destaining

solutions were prepared manually but according to

the protocols given by BioRad. Three different

conditions of PAGE were performed, also following

the protocols by BioRad:

– SDS PAGE this was slightly modified from the

protocol described in our previous paper (Stol-

lenwerk et al. 2010). Briefly, 7.5% Tris–HCl gels

containing no SDS were used. The amount of

SDS necessary to denature the proteins was

present only in the sample and running buffers.

The sample concentrations were adjusted to 0.5 or

1 lg/ll and after addition of the sample buffer

they were heated to 95�C for 5 min. Immediately

afterwards they were loaded onto the gel together

with a molecular weight standard. The loading

volumes were 10 ll for the standard and 20 ll for

each sample, the gel running conditions were set

to 50 V for 5 min and 150 V for 60 min. After

the electrophoresis was complete the gel was

removed, stained with Coomassie blue for 1 h and

destained until the desired background was

reached.

– Native PAGE also here 7.5% Tris–HCl gels

containing no SDS were used. Under native

conditions SDS is neither present in the sample

nor in the running buffers. The samples were not

heated but loaded directly onto the gel after

addition of the sample buffer. All other conditions

were the same as for SDS PAGE.

– Isoelectric focussing (IEF) PAGE IEF gels were

used having a pH range from 4 to 8.5. The sample

concentrations were adjusted to 0.5 or 1 lg/ll.

After addition of the sample buffer they were

loaded directly, without being heated, onto the gel

together with an IEF standard. The loading

volumes were 10 ll for the standard and 20 ll

for each sample. The gel running conditions were

set to 100 V for 60 min, 250 V for 60 min and

500 V for 30 min. After the electrophoresis was

complete the gel was removed, stained with IEF

stain for 1 h and destained until the desired

background was reached.

20. Modelling the reliability of TEM measurements of

nanoparticle size a model ensemble of particles

(glass marbles of various sizes) was photographed to

provide well-defined images allowing particles of

many sizes to be counted and measured. Particle

collectives were counted at different numbers

(n = 1 - 1,000), then recounted with the removal

of a single small particle, a large-sized particle, or

one of each. For each collective, the average size

and the polydispersity index (PDI) were calculated

as described earlier in detail (Stollenwerk et al.

2010).

21. Numbers of nanoparticles in 1 g the concentration of

a nanoparticle colloidal solution was determined by

UV spectrometry at 280 nm. The solution was

adjusted to a concentration of 1 lg/ml (10-6 g/ml)

by adding water. A dilution series was prepared in

which each member was tenfold more dilute, the final

concentration being 10-12 g/ml. From each solution

in the dilution series, a 1 ll drop was allowed to dry

onto a formvar-coated slot copper grid. Each grid was

dipped for 10 s into a 1% solution of uranyl acetate in

30% methanol, then dried under a lamp. Grids were

viewed in a Philips CM 120 electron microscope, and

a grid selected for counting which bore easily

countable populations of nanoparticles. The dilution

factor associated with this grid was noted, and was

employed in the subsequent calculations, which were

used to calculate the number of nanoparticles per

milliliter at that concentration of the nanoparticle

solution.

Efficiency calculations

Synthesis yields were calculated on the basis of the amount

of HSA protein used as starting material. UV spectrometric

measurements made immediately after diafiltration of the

purified protein were used to calculate the efficiency of this

step, and similar measurements made immediately after

diafiltration of the HSA-DTPA conjugates or, in some

176 Histochem Cell Biol (2010) 134:171–196

123

cases, the HSA-DTPA-Gd conjugate-chelates, were used to

calculate the efficiency of the synthesis prior to emulsifi-

cation with PLA. The calculations were carried out for

each batch individually and for the conjugation/chelation

step incorporated the assumptions that a single gadolinium

ion was chelated by a DTPA moiety and that there were no

unoccupied DTPA moieties; the absence of non-chelated

gadolinium ions had been previously checked by arsenazo

III analyses.

Storage of the nanoparticles

Nanoparticle preparations were stored in water in the

presence of 0.1% sodium azide, at 4�C.

Results

Monitoring and improving purity

Dialysis was used for early work on batch 1 of series III. Its

efficacy was tested, using cutoffs (100, 500 kDa) well

above the molecular weight of HSA (MW 66,400). Pierce

and AAS (Fig. 1a) and MRI (Fig. 1b) analyses showed that

dialysis of HSA proceeded extremely slowly in water and

that it failed entirely in saline and PBS; use of the higher

cutoff (500 kDa) did not improve the results. In conse-

quence, diafiltration was introduced in late stages of batch

1 and used in all later batches for purification of interme-

diate stage products and of nanoparticles. At the same time,

a suite of analytical techniques was assembled to allow

rapid detection of contaminants and thus monitor progress

of the purification. TLC was the first of these techniques. It

was used to check both conjugates and nanoparticles. It

confirmed earlier Pierce, AAS and MRI results in series II

and batch 1 of series III, which in some batches indicated

the presence of unexplained excesses of gadolinium and

implied the presence of varying amounts of educts con-

taminating both the conjugates and the nanoparticles. TLC

also aided analysis of intermediate steps during the course

of the reactions which conjugated DTPA to HSA, and then

chelated Gd to the HSA-DTPA conjugates (Fig. 2); it

revealed the presence of major contaminant materials, in

this case non-bound DTPA and non-chelated gadolinium.

Development of the TLC sheets in arsenazo III solution

(see below) showed non-chelated gadolinium present in

the [Gd(NTA)2]Na3 (NTA-Gd) stock solution (Fig. 2a).

Figure 2b demonstrates that the (diafiltered) conjugates were

not contaminated with non-chelated gadolinium (lane 5 is a

positive control). This method is a novel means of visualis-

ing non-chelated gadolinium, and Fig. 2a, b illustrate its

utility. TLC developed by exposure to iodine vapour

revealed impurities in preparations of both conjugates and

nanoparticles. Stabilised HSA was free of contaminants,

whereas HSA-DTPA-Gd conjugates were contaminated by

NTA and/or DTPA prior to diafiltration (Fig. 2c). The large

total amounts of educt present in filtrates (Fig. 2c, d, lanes

4, 5, respectively) show the effectiveness of diafiltration in

purifying the conjugates and the nanoparticles. Diafiltra-

tion, however, altered the conformation of the albumin

(compare lanes 1, 4 in Fig. 2c, and 1, 3 in d). For nano-

particles, Fig. 2e shows a particularly favourable result, in

which the nanoparticles exhibited ideal purity and also

good stabilisation. Figure 2f, in contrast, shows two

nanoparticle batches in which glutaraldehyde fixation

appears to have displaced some DTPA from the conjugates.

Arsenazo III chromogenic analysis provided a second

rapid procedure for monitoring the progress of purification.

Arsenazo III solutions distinguished between chelated and

non-chelated gadolinium: the presence of chelated gado-

linium was indicated by a change of arsenazo III colour

from bright pink to lighter or darker purple, depending on

the concentration of chelated gadolinium present, whereas

a change to green indicated the presence of non-chelated

gadolinium (Fig. 3a). In diafiltered preparations of

Fig. 1 Retention of albumin and albumin-DTPA-Gd conjugates in

dialysis tubing throughout periods 15 min to 50 h. a Dialysis using

both 100 and 500 kDa cutoffs, results analysed by AAS. b Dialysis

using 100 kDa cutoff, results analysed by MRI

Histochem Cell Biol (2010) 134:171–196 177

123

conjugates or nanoparticles, this method showed that non-

chelated gadolinium was not present (Fig. 3b). The arse-

nazo III method allowed the course of gadolinium binding

to HSA, followed by the removal of the bound gadolinium,

to be visualised effectively: Fig. 3c shows an experiment

to assess whether gadolinium binds to HSA which bears no

DTPA moieties. Tubes 2 and 7 in the Figure showed

that the gadolinium did indeed bind to the pure HSA, and

this was confirmed by AAS/Pierce analysis, which mea-

sured *1.7 gadolinium ions per HSA molecule (we con-

sider it likely that these gadoliniums were bound to the two

metal-binding sites of HSA). Further testing with arsenazo

Fig. 2 Analysis of intermediate products by thin layer chromatog-

raphy (TLC). a TLC in the citric acid/citrate system, developed by

immersion in arsenazo III. Lane 1 NTA, lane 2 NTA:Gd stock

solution, lane 3 GdCl3, lane 4 DTPA, lane 5 Magnevist = DTPA-Gd.

The green line shows non-chelated gadolinium. Note that Magnevist

shows no non-chelated gadolinium. b TLC in the citric acid/citrate

system, developed by immersion in arsenazo III. The arsenazo has

stained the TLC sheet lightly pink, except where the presence of large

molecules or particles has hindered its access to the sheet material.

Lane 1 HSA-DTPA-Gd of batch 15, before diafiltration, lane 2 HSA-

DTPA-Gd of batch 15, after diafiltration: retentate (i.e.: containing the

purified conjugate), lane 3 HSA-DTPA-Gd of batch 15, after

diafiltration: filtrate (i.e.: containing the impurities which have been

removed by diafiltration), lane 4 nanoparticles of batch 15, after

glutaraldehyde fixation, lane 5 GdCl3 as positive control. Note that

only lane 5 shows the presence of non-chelated gadolinium, in the

form of a green line at the forward edge (asterisk). c TLC in the citric

acid/citrate system, developed by exposure to iodine vapour, showing

preparations from batch III/11. The iodine has stained the TLC sheet

material yellow, and all amines brown. Lane 1 stabilised HSA, lane 2HSA-DTPA, before diafiltration: note the presence of a dark heart-

shaped spot indicating the presence of DTPA not attached to the HSA,

lane 3 HSA-DTPA-Gd before diafiltration: here also, a prominent

spot indicates the presence of DTPA not attached to the HSA, lane 4HSA-DTPA-Gd, after diafiltration: retentate: we suggest that the faint

spot (beneath the DTPA spot in lane 3) represents DTPA covalently

attached to the HSA. Lane 5 HSA-DTPA-Gd after diafiltration:

filtrate; note that the material in lane 5 was a small sample (*2 ll)

from a large volume (600 ml) of diluted filtrate: in this strongly

diluted sample the moderately intense spot represents a much larger

total of DTPA in the total filtrate. d TLC in the citric acid/citrate

system, developed by exposure to iodine vapour, showing prepara-

tions from batch III/13. Lane 1 stabilised HSA, lane 2 HSA-DTPA-

Gd before diafiltration: note the intense heart-shaped band indicating

the presence of DTPA not bound to the HSA. Note also an extended

faint spot (in blue oval) which we consider to represent NTA, lane 3HSA-DTPA-Gd, after diafiltration: retentate, lane 4 HSA-DTPA-Gd,

after diafiltration: filtrate, in this strongly diluted sample the

moderately intense spot represents a much larger total of DTPA in

the total filtrate. e TLC in the citric acid/citrate system, developed by

exposure to iodine vapour, showing preparations from batch III/3.

Lane 1 HSA-DTPA-Gd, after diafiltration: retentate, lane 2 stabilised

nanoparticles after diafiltration: retentate, lane 3 stabilised nanopar-

ticles after diafiltration: filtrate, lane 4 stabilised nanoparticles not

diafiltered (and the result of an experiment in which HSA:PLA was

emulsified in the ratio 1:1). Note that in lanes 2, 4, the nanoparticles

occupy a round spot at the origin, indicating the absence of

conformational isoforms, i.e.: an ideal result—these nanoparticles

are shown again in Fig. 4a, lanes 3–4. f TLC in the citric acid/citrate

system, developed by exposure to iodine vapour, showing prepara-

tions from batches III/14 (lanes 1, 2) and III/15 (lanes 3, 4). Lane 1batch 14 nanoparticles before glutaraldehyde fixation, Lane 2 batch

14 nanoparticles after glutaraldehyde fixation: note the presence of

non-bound DTPA which is not evident in lane 1, lane 3 batch 15

nanoparticles before glutaraldehyde fixation, Lane 4 batch 15

nanoparticles after glutaraldehyde fixation: note again the presence

of non-bound DTPA which is not evident in lane 3

c

178 Histochem Cell Biol (2010) 134:171–196

123

III then showed that it was possible to remove all the bound

gadolinium by adequate diafiltration (tubes 3–5 in the

Figure), and that this was more effective in the presence of

citrate ions (tubes 6–10 in the Figure); this was confirmed

by AAS. A further example of the use of arsenazo III

resulted from our query whether our NTA stock solution

contained non-chelated gadolinium. Figure 3d shows that

the stock NTA-Gd solution did contain non-chelated gad-

olinium (tube 2 in the Figure), and that increasing the

proportion of NTA only gradually chelated more of the

gadolinium (tubes 3–7 in the Figure), and all the gadolinium

was chelated only when high molar ratios (NTA:Gd [ 5:1)

were used (tubes 8, 9 in the Figure).

Polyacrylamide gel electrophoresis (PAGE)

PAGE was used to assess mechanical stability of the nano-

particles under a range of various preparative procedures,

and was also used to assess the charge:mass properties of

both nanoparticles and conjugates. To assess the mechanical

stability of individual nanoparticles, nanoparticle colloidal

solutions were placed on SDS–PAGE gels. The particles are

larger than the pores within the gel, so that cohesive (stable)

particles should not enter the gel: they remained at the gel

origin. Macromolecules such as HSA and HSA-DTPA-Gd,

being much smaller, entered the pores within the gel and

migrated through the gel according to their mass and charge

properties (Fig. 4a, lanes 2, 6, 7). The presence of SDS in the

gel aided the disintegration of unstable particles, reducing

them to their macromolecular constituents (Fig. 4a, lanes 3,

4, 8, 9). In certain nanoparticle batches, some mechanically

stable particles were present, and these did not enter the gel

(Fig. 4a, lanes 3, 4); less stable nanoparticles entered the gel

and disintegrated there [note that the batch of nanoparticles

shown in lanes 3, 4 also exhibited favourable properties in

the TLC assay (Fig. 2e, lanes 2, 4)]. In other batches of

nanoparticles, none of the particles exhibited adequate

mechanical cohesion to resist entering the gel (Fig. 4a, lanes

8, 9). Batches of nanoparticles which were not only glutar-

aldehyde fixed (generating Schiff bases between the protein

Fig. 3 Analysis of gadolinium chelation by arsenazo III staining.

a The three colours of arsenazo III, photographed from a 12-well

plastic dish. At top, the pink colour characterises the arsenazo III

solution in which no gadolinium is present. At centre, the greencolour shows that gadolinium is present in non-chelated (‘‘free’’)

form. At bottom, the purple colour shows that gadolinium is present,

but is fully chelated: no free gadolinium ions are present. Since

arsenazo colour depends on pH of the solution, it should be noted that

the colours shown here were recorded at pH 4.0. Note that b, c have

been lightly retouched to remove background details. b Conjugates

and nanoparticles. Tube 1 stock NTA-Gd, Tube 2 conjugate 5 min

after addition of stock NTA-Gd, Tube 3 conjugate after 24 h at 4�C

before diafiltration, Tube 4 conjugate after diafiltration: retentate,

Tube 5 pure arsenazo III solution, Tube 6 glutaraldehyde-fixed

nanoparticles before diafiltration, Tube 7 glutaraldehyde-fixed nano-

particles after diafiltration: retentate, Tube 8 glutaraldehyde-fixed

nanoparticles after diafiltration: filtrate. c Experiment testing potential

binding of gadolinium ions to HSA in pure form (without DTPA

attached). The five tubes (1–5) at left show (1) arsenazo III in pure

solution; (2) HSA plus gadolinium prior to diafiltration; (3) HSA plus

gadolinium diafiltered with 150 ml water; (4) HSA plus Gd diafiltered

with 300 ml water; (5) HSA plus gadolinium diafiltered with 450 ml

water. Tubes 6–10 show the same experiment as tubes 1–5, except

that citrate ions are present at a concentration of 0.1 M. d Nine

Eppendorff tubes containing arsenazo III solution (pH 4.0) to which

different combinations of NTA-Gd have been added. At left (1) is

arsenazo III solution without NTA-Gd; the following eight tubes (2–

9), from left to right, contain NTA-Gd in different proportions. Tube 2

contains NTA:Gd at 2:1 ratio, each subsequent tube is a further

increase in this ratio, until the tube at right (number 9) is reached,

which contains 8 NTAs per 1 Gd. Note the development of the colour

from green (tube 2) through blue (tubes 5, 6) to purple (tubes 8, 9),

indicating that an excess of NTA in the ratio of 7:1 or 8:1 is necessary

to ensure that all the gadolinium ions in the solution are chelated

Histochem Cell Biol (2010) 134:171–196 179

123

Fig. 4 Analysis of intermediate products by polyacrylamide gel

electrophoresis (PAGE). a SDS–PAGE of stabilised HSA (lane 6), of

conjugates after diafiltration (lanes 2, 7) and of glutaraldehyde-fixed

nanoparticles after diafiltration (lanes 3, 4, 8, 9). (Note that the particles

shown here in lanes 3, 4 are shown again in Fig. 2e, lanes 2, 4). The

nanoparticles in lanes 8 and 9 break up completely under the denaturing

effect of SDS, only a small proportion of the nanoparticles in lanes 3 and4 (red arrows) remain at the origin. Lane 1 shows the molecular weights

(kDa) obtained by running standards in the same gel. No differences

between conjugates and nanoparticles are visible. b SDS–PAGE of

nanoparticles fixed with different concentrations of glutaraldehyde

(lanes 1–5 1–5% v/v) and quenched with 1% (v/v) NH4Cl, and also

nanoparticles fixed with the same concentrations of glutaraldehyde

(lanes 6–10 1–5%) but reduced with NaCNBH3. A better stability is

reached when the nanoparticles are treated with NH4Cl but they are also

more likely to form visible aggregates (data not shown). (No molecular

weight standard was applied to this gel but interpretation of the masses

is according to the standard shown in a.) c Native-PAGE of stabilised

HSA (lanes 1, 6), of HSA/DTPA after diafiltration (lane 2), of conjugate

after diafiltration (lane 3), of non-glutaraldehyde-fixed nanoparticles

before (lane 4) and after (lane 5) diafiltration and of glutaraldehyde-

fixed nanoparticles after diafiltration (lanes 7, 8). Under native

conditions the molecules migrate through the gel according to their

mass–charge ratio; also here no difference between conjugates and

nanoparticles is visible. (No molecular weight standard is available

for native gel conditions.) d shows SDS–PAGE of stabilised HSA

(lane 6), of conjugate after diafiltration (lane 7), of glutaraldehyde-fixed

nanoparticles (lanes 2, 8), of glutaraldehyde-fixed nanoparticles

PEGylated with PEG 5 kDa (lane 3), with PEG 10 kDa (lane 4) and

with PEG 20 kDa (lane 5), and of glutaraldehyde-fixed nanoparticles

targeted with LEA lectin (lane 9). Lane 1 shows the molecular weight

(kDa) obtained by running standards in the same gel. A clear difference

is visible between PEGylated and non-PEGylated nanoparticles.

PEGylated nanoparticles do not break up to the same extent as non-

PEGylated nanoparticles, and the degree of breakup decreases as the

length of the PEG chains increases (lanes 3–5). e shows IEF-PAGE of

non-stabilised (lane 2) and of stabilised HSA (lane 3), of series IIIconjugate after diafiltration (lane 4), of series III glutaraldehyde-fixed

nanoparticles after diafiltration (lanes 5–7) and of series II lyophilised

nanoparticles (lanes 8–10). HSA clearly shows its natural isoelectric

point (pI) of about 4.7 but no difference in the pI is visible between

conjugates and nanoparticles. Lane 1 shows the pI values obtained from

running standards in the same gel, these bands are blocked out where

they overlapped with lane 2

180 Histochem Cell Biol (2010) 134:171–196

123

constituents) but were also ‘‘quenched’’ with ammonium

chloride, or reduced to covalency by reaction with sodium

cyanoborohydride, contained significant proportions of

nanoparticles with mechanical stability sufficient to resist

entering the gel (Fig. 4b). The ‘‘quenched’’ particles con-

tained large majorities of stable nanoparticles, and the

minorities of particles that disintegrated and entered the gel

produced only weak bands representing macromolecular

constituents (i.e.: disintegration products) migrating through

the gel (Fig. 4b, lanes 1–5). In batches which were reduced

to covalency, the proportions of mechanically cohesive

particles were much smaller and the bands representing

particle disintegration were much more prominent (Fig. 4b,

lanes 6–10). Another form of PAGE, ‘‘native PAGE’’, in

which no SDS detergent is present and sample migration

depends on charge:mass relationships, showed that no dif-

ference could be observed between the HSA-DTPA-Gd

conjugates and nanoparticle samples (Fig. 4c), thus dem-

onstrating that the major component of the nanoparticles is

indeed HSA-DTPA-Gd. Further PAGE (SDS–PAGE)

analyses showed that ammonium chloride quenching and

covalent reduction were not the only methods that protect

nanoparticles against mechanical disintegration: PEGyla-

tion of glutaraldehyde-fixed nanoparticles provided an

additional stabilisation which was clearly seen in SDS–

PAGE (Fig. 4d), and which increased as the mass of the PEG

chains increased (compare lanes 3, 4, 5 in Fig. 4d). On the

other hand, LEA-derivatized nanoparticles showed little

mechanical stability and disintegrated in SDS–PAGE

(Fig. 4d, lane 9). A further form of PAGE, isoelectric

focussing (IEF-PAGE), provided information confirming

the variability of the HSA-DTPA conjugates: a clear pI of

4.7 was seen for HSA, with well-defined, sharp bands

(Fig. 4e, lanes 2, 3). In contrast, conjugates and nanoparti-

cles produced smears in the range pH 5.5–7.5, indicating the

presence of numerous charged variants within the molecular

and particle populations (lanes 4–10, Fig. 4e); we attributed

these to the varying numbers of DTPA-Gd chelates attached

to the HSA.

Size and size standardisation

HSA-DTPA-Gd conjugates

After removal of all ligands from fraction V HSA, and

their replacement by caprylate as single major ligand, the

HSA molecules assumed a standard and highly repro-

ducible set of properties, most easily demonstrated by

PCS as a narrow distribution of particle size close to

10 nm hydrodynamic diameter (Fig. 5a). The size

increased to 14 nm hydrodynamic diameter after attach-

ment of the DTPA groups (Fig. 5b) and after chela-

tion of the gadolinium ions, but remained uniform

(Fig. 5c); both the size and the variability of the protein

remained similar throughout these different stages of the

synthesis. Diafiltration of the conjugates, and emulsifi-

cation of conjugates with PLA, led to some aggregation

(Fig. 5d).

Reliability and validation of single particle counts

The number of particles that should be counted in order to

obtain reliable valid PDI values was investigated by use of

a simple glass marble model (Fig. 6a, b). The sensitivity to

omission of a single particle from the ensemble was high

for small ensembles, for example omitting a single rela-

tively very small particle from an ensemble of size n = 29

(Fig. 6a) altered PDI from 6.168 to 5.96 (-3.37%);

omission of a single large particle changed PDI from 6.168

to 6.595 (?6.92%). Removing a single particle from larger

ensembles altered the PDI values obtained by smaller

amounts, the exact value depending on the relative size of

the particle removed. Removal of a very small particle

resulted in a divergence of the PDI value by \ 1% if 100

particles were counted; removal of a very large particle

resulted in a divergence of \ 1% only after [ 500 particles

had been counted (Fig. 6b). The sample size n = 500 was

therefore selected as the minimum to obtain reliable PDI

values; a sample of nanoparticles of this population size is

shown in Fig. 7a.

Characterisation of the nanoparticles’ size

Starting with homogenous HSA-DTPA-Gd conjugates,

emulsification with PLA produced a turbid solution

containing aggregates of various sizes. Removal of these

aggregates by filtration through paper filters resulted in

populations of nanoparticles which had PDI values in the

range 1.2–1.6, and generally *1.4–1.5. Aggregation also

occurred after later steps in the synthesis, including

diafiltration of fixed or unfixed nanoparticles (Fig. 5e, f).

The almost uniform size of the filtered nanoparticles was

evident in TEM images (Fig. 7a). In most batches the

nanoparticles had average diameters close to 18 nm

(Fig. 7b). The uniform nanoparticle size was reflected in

PDI values in the range 1.2–1.6 (Fig. 7c). In both their

nanoparticle size and the size variability characteristics,

these batches represented a clear improvement on the

previous series II of nanoparticles (Stollenwerk et al.

2010, Fig. 6).

Numbers of nanoparticles per gram

For batch 15, *500 nanoparticles were counted in a single

electron microscopic field of view at 88,0009 (Fig. 7a),

the field of view covering an area *1.7 lm2. The 0.2 ll

Histochem Cell Biol (2010) 134:171–196 181

123

drop of nanoparticle solution applied to the grid had dried

down to cover an area *1 mm2 = 106 lm2) so it had

contained approximately [500 9 106]/1.7 = 2.94 9 108

nanoparticles. Noting our definition of 6.022 9 1023

nanoparticles as representing 1 mol of nanoparticles, we

find that the 0.2 ll drop had contained [2.94 9 108/6.022 9

1023] = 4.88 9 10-16 mol of nanoparticles, which we

rounded to 5 9 10-16 mol.

Numbers of conjugates per nanoparticle

The measured concentration of HSA in the nanoparticle

solution of batch 15, used to prepare Fig. 7a, was

1.5 9 10-5 mol/ml, and that solution was diluted by a factor

106 prior to application to the grid, so the 0.2 ll drop of this

solution had contained 3.01 9 10-15 mol of HSA. The

solution contained*3 9 10-15 mol of HSA arranged in the

Fig. 5 Follow-up of our nanoparticle synthesis shown by volume-

weighted PCS read-outs, the numbers given at each peak represent its

mean size, its standard deviation and the percentage of the total

particle population of that peak. a Cleaned and stabilised HSA,

b shows HSA/DTPA (not diafiltered), c shows HSA-DTPA-Gd before

diafiltration, d shows HSA-DTPA-Gd after diafiltration, note that

diafiltration causes mechanical stress which disturbs the protein

conformation and leads to minor aggregation, e shows nanoparticles

after emulsification with PLA, not yet fixed with glutaraldehyde and

not diafiltered, note that emulsification leads to strong formation of

aggregates, f shows nanoparticles, fixed with glutaraldehyde and

diafiltered

182 Histochem Cell Biol (2010) 134:171–196

123

form of *5 9 10-16 nanoparticles, indicating that each

mole of nanoparticles contained [3 9 10-15/5 9 10-16] =

6 mol of HSA. Thus each 18 nm nanoparticle comprised six

molecules of HSA-DTPA-Gd (plus a small amount of PLA).

Further calculations become possible when the packing

density of the nanoparticles is known (see below).

Packing density of the nanoparticles

Nanoparticles in batch 15 were (on average) 18 nm in

diameter, and we calculate the packing density for a

nanoparticle of this size and containing six conjugate

molecules. The nanoparticle volume was [(4/

3)p(9)3] nm3 = 3,054 nm3. Taking the volume of HSA to

be 90 nm3 (Stollenwerk et al. 2010; Sugio et al. 1999), a

nanoparticle containing six HSA molecules (total vol-

ume = 540 nm3) has a packing density (540/

3,054) 9 100% = 17.7%. This calculation ignores, how-

ever, the DTPA-Gd chelates attached to the nanoparticles,

which increase their volume; this will be considered more

closely in the ‘‘Discussion’’. Taking the packing density

value based simply on HSA, a nanoparticle of 20 nm

diameter, with a volume of 4,188 nm3, of which 17.7%

(741 nm3) consists of protein material, would contain (741/

90) = 8.2 conjugate molecules. In our series III batches,

nanoparticle sizes are between 18 and 22 nm diameter, so

our evidence indicates that the nanoparticles each consist

of 6–10 HSA-DTPA-Gd conjugate molecules, together

with a small amount of PLA (see below).

Electrical charge on the conjugates and nanoparticles

Surface charge measured as zeta potential

HSA-DTPA-Gd conjugates carried a small positive charge,

between 1 and 4 mV. Nanoparticles formed from these

conjugates carried a larger and negative charge, -30 to

-40 mV, which, however, disappeared after the nanoparticle

Fig. 6 Glass marble model experiment to determine the minimum

counts necessary to generate reliable size and PDI data from TEM

images of the nanoparticles. a Glass marble collectives to demonstrate

the concept underlying the PDI. At left, 29 marbles form a population

of various sizes, the variability amongst the marbles being summa-

rised by the large value for PDI: 6.16. At right, 17 marbles form a

population of marbles of closely similar size, the uniformity of the

sizes being summarised by the small value for PDI: 1.15. For marbles

of identical size, PDI = 1.0. A small marble (red asterisk) or a large

marble (blue asterisk) were removed from the population at left

during counting experiments, see b. b Data from repeated counts of

the marble population shown in a. During one of the many counting

sessions, one large marble, one small marble, or one large and one

small marble, were removed, thus altering the PDI. After repeated

counting sessions the alteration (carried out once only) became

insignificant in its effect on PDI, and the calculated PDI therefore

converges on the original value. The convergence is seen as the

‘‘error’’ curves approach the zero line. For each curve, counts of

increasing size were made, the largest being 579 counts. Pink line:

removal of a single large marble; green line: removal of one small and

one large marble; blue line: removal of a single small marble. The red

number placed against each line indicates the number of counts

necessary to achieve an ‘‘error rate’’ less than 1% for this type of

‘‘variation’’

Histochem Cell Biol (2010) 134:171–196 183

123

solutions were lyophilised. Nanoparticles targeted by

attachment of the LEA lectin also carried significant neg-

ative charge (-20 to -30 mV).

Density of the conjugates and nanoparticles

Conjugates and nanoparticles were assayed by pyknometry

in concentrations ranging from 5 to 25 mg/ml, the upper

limit being set by the maximum nanoparticle concen-

tra-tion available after diafiltration; HSA was assayed up to

30 mg/ml. All results obtained from solutions at concen-

trations below 10 mg/ml were below the detection limit of

our pyknometry. The data obtained from solutions above

20 mg/ml showed a trend: based on the remaining small

number of measurements, the density of stabilised HSA

was 1.86 g/cm3, that of the HSA-DTPA conjugate 1.58

g/cm3, that of the HSA-DTPA-Gd conjugate 1.75 g/cm3

and that of the nanoparticles 1.38 g/cm3.

Structural stability of the nanoparticles

As noted above, treatment of the nanoparticles with glu-

taraldehyde, or with glutaraldehyde followed by exposure

to ammonium ions, or to glutaraldehyde followed by

reduction to covalency, all enhanced the mechanical sta-

bility of the nanoparticles. PEGylation with 20 kDa PEG

chains also protected the nanoparticles from breaking up in

polyacrylamide gels.

Fig. 7 Size analysis of the nanoparticles. a A negative contrast

transmission electron microscopical image of nanoparticles of batch

III/15, following glutaraldehyde fixation and diafiltration. Original

magnification 988,000. Calibration bar 200 nm. b A dot diagram of

the nanoparticle sizes in series III batches. Each dot represents a

series III batch and is denoted by the batch number. The scale is the

same as that used in our previous paper (Stollenwerk et al. 2010). c A

dot diagram of the PDI values in series III batches. Each dotrepresents a series III batch and is denoted by the batch number. The

scale is the same as that used in our previous paper (Stollenwerk et al.

2010)

184 Histochem Cell Biol (2010) 134:171–196

123

Conformation of protein components

in the nanoparticles

After glutaraldehyde fixation and ammonium chloride

quenching, the nanoparticles were unaltered in size

(20–30 nm diameter), and retained the antigenic epitopes

characteristic for human serum albumin (Fig. 8).

Number of gadolinium ions in the nanoparticles

Characterisation of the HSA-DTPA-Gd conjugates

The conjugates were checked for purity, in particular for the

absence of contaminating educts such as non-chelated

gadolinium, as described above. Then the Gd:HSA ratios

were determined by Pierce/AAS assays. As shown in Fig. 9,

early batches in this series showed relatively low Gd:HSA

ratios, but later batches had higher ratios and the final bat-

ches tended to cluster in the Gd:HSA = 15–20 range. An

experiment carried out to assess how many gadolinium ions

could bind to each molecule of pure HSA (not bearing

DTPA) gave the result 1.7 (there are two metal-binding sites

in each HSA molecule); adequate diafiltration removed

these gadolinium ions from the HSA molecules (Fig. 3),

and since adequate diafiltration was routinely applied, this

type of non-specific binding was ignored during the pre-

paration of all later batches. A nanoparticle with six conju-

gate molecules has packing density *17%, mass 461 kDa

and relaxivity r1 = 0.83 9 106/Ms. It carries 114 Gd ions

each with relaxivity *7,280/Ms.

Relaxivities of the conjugates and nanoparticles

The conjugates in this series III exhibited higher relaxivi-

ties than in the previous series II nanoparticles. After an

initial learning period, comprising the first five batches of

series III, we obtained conjugates with r1 values routinely

exceeding 3 ml/mg s, and clustered between 3 and 4 ml/

mg s; their r2 values all exceeded 4 ml/mg s, clustering

between 4.5 and 5.5 ml/mg s (Fig. 10a). Similarly, the

nanoparticles, after the first five batches, exhibited r1

Fig. 9 Dot diagram of Gd:HSA ratios for nanoparticle batches in

series III. Each dot represents a single batch, denoted by its series IIIbatch number. Values for LEA-targeted nanoparticles are shown in

red. Batch 5 was derivatized with too small an amount of gadolinium.

Its low Gd:HSA ratio resulted in a low relaxivity (compare Fig. 10);

this batch is shown here because it helps exemplify the relation

between Gd:HSA ratio and relaxivity (Fig. 11). a Data for the HSA-

DTPA-Gd conjugates, b data for the nanoparticles

Fig. 8 Series III batch six nanoparticles after glutaraldehyde fixation

and ‘‘quenching’’ with ammonium chloride. Original magnification

940,000. The sample was acquired from a PAGE application well

after the gel had been run, and essentially no nanoparticles had

entered the gel. The sample on the grid was incubated for anti-HSA

immunohistochemistry, with the second antibody bearing gold

particles

Histochem Cell Biol (2010) 134:171–196 185

123

values exceeding 1.5 ml/mg s, clustering between 1.5 and

2.0 ml/mg s; their r2 values all exceeded 2.5 ml/mg s,

clustering close to 2.6 ml/mg s (Fig. 10b).

The analytical data presented above showed that the

nanoparticles were of similar sizes and chemistry in bat-

ches 2–15 of series III. The coefficient of determination,

R2 *81%, seen in the regression plot (Fig. 11) between

Gd:HSA and relaxivity, was therefore an indication that

nanoparticle relaxivity depended directly on the number of

gadolinium ions present on the nanoparticles. The average

number of gadolinium attached to the HSA conjugate

molecules of the series III nanoparticles was 19 (Fig. 9a),

so the conjugate molecules had average molecular weight

([393.35 ? 157.25 = ] 550.6 9 19) ? 66,400 = 76,861

Da = *76,9 kDa. Consisting of six conjugate molecules,

a nanoparticle with diameter 18 nm therefore weighed

(6 9 76.9) = 461 kDa = 0.461 MDa. A typical value for

relaxivity r1 was 1.8 ml/mg s (Fig. 10), so the relaxivity

for the series III nanoparticles was close to r1 =

(0.461 9 1.8) = 0.830 9 106 /Ms. A similar calculation,

taking typical R2 values for series III as 2.6, indicates that

the nanoparticles had r2 = 1.199 9 106 /Ms.

Toxicological assessment of the nanoparticles

Cell viability remained above 90% at nanoparticle doses of

2 mg/ml, as measured by ATP content, and no membrane

damage was seen, as assessed by release of the cytoplasmic

enzyme lactate dehydrogenase; some loss in viability was

seen at 5 mg/ml (Fig. 12a); LDH release data showed a

similar pattern (Fig. 12b). Three of the four batches tested

did not cause hemolysis even at 5 mg/ml nanoparticle

concentration, and the fourth batch caused a maximum of

5% hemolysis at 5 mg/ml nanoparticle concentration

(Fig. 12c). Plasmatic coagulation was assessed by quanti-

fication of prothrombin F1 ? 2 fragments, which arise

during activation of the cascade. The level of the fragments

increased 12 times upon treatment with the positive control

(5 mg/ml kaolin) but was around 80% for the particles;

platelets showed aggregation upon treatment with the

positive control 20 lM ADP. In the negative control and in

the nanoparticle samples up to 5 mg/ml, pseudopod for-

mation of thrombocytes was seen occasionally, but no

aggregation of platelets occurred (Fig. 13).

Fig. 10 Dot diagram of HSA-DTPA-Gd conjugate and nanoparticle

relaxivities in series III. a Conjugate relaxivities are shown in the

form ‘‘ml/mg s’’; the red entries show the relaxivity r1, the blueentries show relaxivity r2. Due to improvements in our synthesis

technique, this diagram extends further to the right than the equivalent

diagram in our previous paper (Stollenwerk et al. 2010). The value for

batch 11 (off-scale in this diagram) was 6.11 ml/mg s. b Nanoparticle

relaxivities are shown in the form ‘‘ml/mg s’’; the red entries show

the relaxivity r1, the blue entries show relaxivity r2. Here also the

scale extends further to the right than in our previous paper

Fig. 11 This plot relates the Gd:HSA ratios of most series III batches

to their r1 relaxivities (batches 9, 10 were not derivatized with

gadolinium, and batch 1 was not fully purified). About 81% of the

variation in the response variable (r1 relaxivity) can be explained by

the explanatory variable (Gd:HSA ratio)

186 Histochem Cell Biol (2010) 134:171–196

123

Attaching targeting groups to the nanoparticles

Four of the fifteen batches (numbers 1, 2, 4, 11) were

derivatized by attaching the LEA lectin to the nanoparti-

cles. Serial dilution experiments allowed the minimum

concentration of LEA-bearing nanoparticles to be identi-

fied at which the particles agglutinated fresh human blood;

the minimum necessary LEA-nanoparticle concentration

was then compared with the minimum necessary LEA

concentration (Fig. 14). The results, shown in Table 1,

show that LEA-bearing nanoparticles were between 0.669

and 20.29 as effective as pure LEA in hemagglutination of

fresh human blood, with most of the ratios lying in the

range 69–129.

PEGylation of the nanoparticles

As shown by PAGE analysis, glutaraldehyde-fixed nano-

particles (Fig. 4d, lane 2) showed some resistance to

mechanical disruption. PEGylated nanoparticles showed

slightly more resistance if the PEG chains were of 5 kDa

mass (Fig. 4d, lane 3), significantly more resistance with

PEG chains of 10 kDa mass (lane 4), and strong resistance

to disruption with PEG chains of 20 kDa mass (lane 5).

Yield efficiencies of nanoparticle syntheses

Yields were calculated on the basis of HSA content, and

expressed as per cent fraction of the amount of HSA pro-

tein used as starting material for any particular batch.

Figure 15 plots two of the major synthetic steps in the

synthesis protocol, namely the cleaning and stabilisation of

the HSA protein, and the conjugation of DTPA/chelation of

Fig. 12 Toxicological data for series III batches 9 and 11. In batch 9,

the nanoparticles were PEGylated, and control nanoparticles were left

non-PEGylated. In batch 11, the nanoparticles were targeted with

LEA, and control particles were left untargeted. The plots show

a viability data, b LDH release data and c hemolysis data; in each

case no toxicity is evident at concentrations of any nanoparticles up to

2 mg/ml. At 5 mg/ml toxic effects begin to become visible

Fig. 13 Platelet aggregation test. a The HSA-based nanoparticles

cause no aggregation of the platelets; rarely a slight activation

(pseudopod formation) of a few platelets is seen (arrow). Calibrationbar 20 lm. b In the positive control, platelet aggregation is common

(arrows). Calibration bar 20 lm

Histochem Cell Biol (2010) 134:171–196 187

123

gadolinium to the HSA. The Figure shows that cleaning of

the protein was associated with relatively high loss rates,

typically close to 30%. The approximately 65% efficiency

of this step was the major determinant of overall efficiency:

the further *10% loss during conjugation/chelation, with

their associated purification procedures, did not involve a

critical reduction in overall synthesis efficiency. A learning

effect is evident in Fig. 15: only after the first five batches

did the yields become predictable; this is evidence for the

complex nature of the conjugation/chelation step, with its

numerous—and sometimes critical—variable parameters.

Discussion

The work described in this paper was a further step towards

optimising the HSA-based nanoparticles we described

previously (Stollenwerk et al. 2010); the improvements in

nanoparticle standardisation, purity and mechanical sta-

bility will be considered here. The question which chelator

to use for attachment of gadolinium will be left to a later

paper: for this work, the linear chelator DTPA was used as

previously. Suspecting that the intra- and interbatch vari-

ation seen in series II was partly due to contamination of

the HSA-DTPA-Gd conjugates by educts, we used protein

analyses and AAS to measure Gd concentration, PAGE to

demonstrate stability, mass and charge:mass properties of

the HSA protein and the HSA conjugates, TLC to visualise

several types of impurities, MRI to measure nanoparticle

relaxivity, PCS and TEM to determine nanoparticle size

and zeta measurements to determine surface charge. These

analyses showed that dialysis, as used in series II, had

indeed failed to remove most educts, leaving for example

single HSA-DTPA-Gd molecules amongst the nanoparticle

products; the presence of remaining educts skewed

size distributions, because singlets, duplexes and higher

aggregates of HSA and HSA-DTPA-Gd conjugates

remained in the samples and were measured together with

Fig. 14 Hemagglutination data from batches 2, 4, comparing LEA

with LEA-bearing nanoparticles. The figure has been lightly

retouched to remove background details. a A positive control series

in which LEA in pure form was applied, b, c two experimental series

in which the LEA was attached to nanoparticles. a 200 ll fresh

human blood mixed with 200 ll LEA solution, as a dilution series.

LEA concentration in tube 1 1 mg/ml, 2 500 lg/ml, 3 250 lg/ml, 4125 lg/ml, 5 62.5 lg/ml, 6 31.25 lg/ml, 7 15.63 lg/ml, 8 7.82 lg/

ml, 9 negative control, PBS only. Red asterisk at tube 6 shows the

minimal LEA concentration causing hemagglutination. b 200 ll fresh

human blood mixed with 200 ll LEA-bearing nanoparticles in

colloidal solution, as a dilution series. Nanoparticle-LEA concentra-

tion in tube 1 1 mg/ml, 2 500 lg/ml, 3 250 lg/ml, 4 125 lg/ml, 562.5 lg/ml, 6 31.25 lg/ml, 7 15.63 lg/ml, 8 7.82 lg/ml. Red asteriskat tube 6 shows the minimal LEA concentration causing hemagglu-

tination. c 200 ll fresh human blood mixed with 200 ll LEA-bearing

nanoparticles in colloidal solution, as a dilution series. Nanoparticle-

LEA concentration in tube 1 200 lg/ml, 2 100 lg/ml, 3 50 lg/ml, 425 lg/ml, 5 12.5 lg/ml, 6 6.25 lg/ml, 7 3.13 lg/ml, 8 1.56 lg/ml.

Red asterisk at tube 3 shows the minimal LEA concentration causing

hemagglutination

Table 1 The results of hemagglutination assays to determine the efficacy of LEA-bearing nanoparticles in agglutination of fresh human blood

Nanoparticle

batch details

Minimal nanoparticle concentration

causing hemagglutination (M)

Minimal LEA concentration

causing hemagglutination (M)

LEA:Nanoparticle ratio, to compare molar

efficacy in agglutinating human blood

Batch series III, #1 0.8–1.9 9 10-11 4.5–9.0 9 10-11 2.4–11.25

Batch series III, #2 2.1 9 10-11 1.8 9 10-10 8.6

Batch series III, #4 8.9 9 10-12 1.8 9 10-10 20.2

Batch series III, #4 2.86 9 10-11 1.8 9 10-10 6.3

Batch series III, #4 1.43 9 10-11 1.8 9 10-10 12.6

Batch series III, #11 5.4 9 10-10 3.57 9 10-10 0.66

188 Histochem Cell Biol (2010) 134:171–196

123

the nanoparticles. This partial failure of purification

explained the strong variability of the nanoparticles syn-

thesized in series II. Whereas dialysis failed to remove

these contaminants, diafiltration did so efficiently. By

repeated cycles of diafiltration it was possible to obtain

nanoparticles in which all the Gd was present as DTPA

chelates conjugated to the nanoparticles. The first major

alteration to our previous synthesis protocol was therefore

to replace dialysis with diafiltration.

The second major alteration to our previous synthesis

protocol was to de-fat and stabilise the HSA prior to pre-

paring HSA-DTPA conjugates. This was motivated by our

suspicion that variability in fraction V HSA played a major

role in causing variability in intra- and interbatch physi-

cochemical properties. The high interbatch variability of

HSA in fraction V (Cohn) was documented 40 years ago

(Chen 1967; Foster et al. 1965). We suspected that HSA

variability affected its properties as surfactant for PLA,

causing unpredictable variations in nanoparticle size. Pre-

viously we had used HSA as obtained from the commercial

supplier to prepare the nanoparticles (series I, series II),

finding variable composition (Stollenwerk et al. 2010). We

now followed Chen’s protocol to de-fat and purify fraction

V HSA: this reduced yield significantly, causing loss of

about one-third of the starting HSA. However, the resulting

HSA and HSA-DTPA were homogenous in size. These

HSA molecules were so reproducible that we could use

them as an informal means of calibrating PCS data. The

nanoparticles made from them were also reliably homo-

genous in size. In order to obtain nanoparticles consisting

of correctly conformed HSA molecules, we added octa-

noate (‘‘caprylate’’), because albumin must bind at least

one of its major ligands in order to be stable in its correct

configuration (Carter and Ho 1994; Spector 1975; Stewart

et al. 2003; Sugio et al. 1999). The resulting albumin was

highly homogenous, chemically well defined, pure, and

consisted of HSA molecules in their correct native con-

formation, as shown by immunohistochemistry.

To overcome further challenges associated with nano-

particle purity and size, we established an on-site suite of

methods providing rapid and sensitive assessment of

whether a synthesis was ‘‘on course’’. TLC, PCS, TEM and

arsenazo III were the techniques used most often on a

routine basis. This series of evaluations gave a much more

comprehensive picture of the composition of the compo-

nents than could a single method. We observed the

importance of comparing different measures of the same

parameter. For example, TEM visualised the nanoparticles

but ignored the aggregates, whereas PCS highlighted the

presence of aggregates but provided only indirect, mod-

elled indications of nanoparticle size (for a full discussion,

see Stollenwerk et al. 2010). Diafiltration caused mechani-

cal stress and led to aggregation; glutaraldehyde fixation

sometimes led to displacement of DTPA moieties from the

nanoparticles. A further example was the assessment of

conjugate and nanoparticle concentrations. UV spectro-

metry was rapid, but its validity for measuring protein

concentrations is doubtful; it is less accurate than Pierce

measurements, which, however, are much slower to per-

form. We noted that Pierce data routinely showed results

about 10% lower than those produced by UV spectrometry.

Both these methods were compared with the results of

drying down conjugate and nanoparticle solutions, and this

underlined the difference between Pierce data and the true

weights: Pierce measures the HSA concentrations and does

not include the DTPA and gadolinium components, and

therefore needs correction by a factor of *1.29. When this

correction is included, the Pierce data are good indicators

of true conjugate concentration. They are good indicators

of true nanoparticle concentration only if the nanoparticles

have been diafiltered, which removes contaminant educts

having a mass approximately equal to that of the nano-

particles themselves. In sum, we repeatedly found that no

single assay could be relied upon to provide a true picture

in the physicochemical characterisation of nanoparticles

and their intermediate synthesis products. Combined

application of the analytical tests at selected stages of

synthesis allowed production of highly purified nanoparti-

cles and facilitated standardisation of the nanoparticle

properties.

TLC was simple to perform, cheap and fast, and could

be used during a synthesis procedure to check for the

Fig. 15 Dot diagram showing efficiency values for the first two steps

in the nanoparticle synthesis. The efficiency of the HSA cleaning and

stabilisation procedure is shown in blue dots, that of the HSA-DTPA-

Gd conjugation and chelation procedure is shown in red dots. Eachdot represents an individual batch synthesis, as shown by the number

associated with the dot

Histochem Cell Biol (2010) 134:171–196 189

123

presence of impurities. Also, TLC sheets were easily

developed by use of a range of different staining methods,

each highlighting one particular aspect of the nanoparticle

chemistry. We used iodine to show all types of amino

groups, ninhydrin to show primary amino groups, fluores-

cence to show conjugated electron systems, and arsenazo

III to show non-chelated gadolinium.

Arsenazo III is a naphthalene derivative with metallo-

chromic indicator properties (Alimarin and Savvin 1966;

Basargin et al. 2000; Rowatt and Williams 1989). It has

been routinely used for detection and assay of lanthanides

in industry, and has also been applied in biomedicine and

nanotechnology (Nagaraja et al. 2006; Soenen et al. 2007;

Magnotti 2008). For example, Nagaraja et al. (2006)

reported the possibility of quantitative assay of Gd(III)

concentration in HSA-DTPA-Gd conjugate samples.

Another protocol (Magnotti 2008) describes the quantita-

tive assay of Gd-chelates in biological samples using

arsenazo III at a lower pH, where Gd binds more strongly

to the arsenazo III than to the chelating agent (DTPA),

resulting in a colour change that can be detected spectro-

photometrically. We found arsenazo III staining a good

complement to TLC analyses, because its striking colour

changes distinguished between free and chelated Gd

(Fig. 3a). We noted thereby that arsenazo III colours are

pH-dependent (Rohwer and Hosten 1997) and therefore

worked always with pH = 4.0; furthermore we were aware

that different gadolinium concentrations result in colour

shifts. The arsenazo analysis of NTA-Gd chelates showed

that they varied significantly in their content of free Gd

(compare Fig. 3d), and that chelation was only complete

for high molar ratios ([5) of NTA to Gd. The incomplete

degree of chelation by NTA suggests that use of this che-

lator might prevent accurate control of conjugation con-

ditions. Using this assay, all nanoparticle batches in series

III were found to be free from non-chelated gadolinium,

though like the conjugates, they varied considerably in

their content of gadolinium chelates. As a first approxi-

mation, the low toxicity of the nanoparticles derives from

this absence of free gadolinium ions.

Gel electrophoresis provided sensitive, stringent tests for

several properties of both conjugates and nanoparticles. It

provided sensitive detection of nanoparticle mechanical

cohesion, with contrasty indication of nanoparticle break-

down. It also visualised variability in chemical properties,

for example variation of DTPA ratios in populations of

conjugates or of nanoparticles were visualised as extended

smears, compared with narrow bands defining homogenous

populations. Electrophoresis also provided evidence for

stabilisation and protection of the nanoparticles in the

presence of PEGylation, showing that 20 kDa PEG chains

were more effective for this than were 5 or 10 kDa chains.

Electrophoresis also indicated that nanoparticles bearing

LEA lectin targeting groups exhibited less mechanical

stability than non-targeted nanoparticles.

Analytical results

Gadolinium content of HSA-DTPA-Gd conjugates

The molar ratios of Gd on the nanoparticles (Gd:HSA)

ranged from 6 to 32, so that series III overall did not differ

from the ratios observed in the previous generation (series

II) of the nanoparticles (Stollenwerk et al. 2010). However,

learning effects in our team are evident in this data (Fig. 9),

see for example that the later batches in series III cluster in

a small range (16–21). In later discussion of these nano-

particles we will use the Gd:NP ratio = 19 as representa-

tive value. Although the Gd:HSA ratio became more

predictable as our skill improved, we consider that the

standardisation of this feature remains one of the more

challenging problems to solve prior to industrial upscaling

of the particles. During synthesis of series III nanoparticles

we applied the gadolinium ions to the HSA-DTPA conju-

gates in the form of NTA-Gd chelates, in order to avoid

any binding of gadolinium ions to pure HSA (i.e.: bearing

no DTPA). However, the arsenazo analysis showed that the

stock solution of NTA-Gd chelate that we used did itself

contain free gadolinium ions, and that these could only be

chelated by adding a large excess (69) of NTA (Fig. 3).

We therefore investigated the results of adding gadolinium

ions in the form of gadolinium chloride, and found that this

resulted in binding of, on average, 1.7 gadolinium ions per

albumin molecule (these albumin molecules bearing no

DTPA). This result is readily explained by the presence of

two metal-binding sites in each HSA molecule, and which

in physiological conditions bind zinc (Stewart et al. 2003;

Andre and Guillaume 2004; Rowe and Bobilya 2000). This

gadolinium binding would be of potential toxicological

significance, because these gadolinium ions could escape

into the tissues and exert significant toxic effects. However,

we found that they could be removed by adequate diafil-

tration, and since in our protocol diafiltration is a routine

purification procedure for HSA-DTPA-Gd conjugates,

these ions would be removed routinely from our nanopar-

ticle preparations. In future work, therefore, we will add

the gadolinium ions to the HSA-DTPA conjugates in the

form of gadolinium chloride, and not as NTA-Gd chelates.

Nanoparticle sizes and size variability

Nanoparticle size and its variability were improved in

series III, when compared with series II. The nanoparticles

in series III were slightly smaller, being generally close to

18 nm diameter rather than 25–30 nm diameter as in series

II (Stollenwerk et al. 2010). The obvious clustering of the

190 Histochem Cell Biol (2010) 134:171–196

123

PDI values to the left of the scale in the dot diagrams for

series III (Fig. 7b, c) indicates a clear improvement over

the diffuse distributions of these values in series II. The

reasons for this improvement lie in the better standardisa-

tion of the HSA-DTPA-Gd conjugates, especially in view

of the indications (see below) that the DTPA-Gd chelate

groups play a significant role in defining the nanoparticle

size. Nonetheless, the series III nanoparticles are not an

entirely uniform particle population. This is illustrated by

the following consideration, which ignores the fact that the

nanoparticles possess not only an outer surface but also a

variety of internal surfaces. The smallest particles, with

diameters 16 nm, have a much larger outer surface area

(1 gm containing 2 9 1018 of these particles has a surface

area of 1,575 m2), whereas the largest particles with

diameters 25 nm have a smaller outer surface area (1 gm

containing 5 9 1017 of these particles has a surface area of

1,026 m2). This difference can be set in perspective by the

fact that the surface area of an adult human being is close

to 1.5 m2, so that the difference in surface area between 1 g

of each of these two sizes of nanoparticles is equal to the

total surface area of *360 adult human beings (weighing a

total of more than 25 tons). Thus, since volume/area ratios

alter rapidly with size in the nanoscale, even the low vari-

ability of these nanoparticles involves dramatic scaling

effects (Gao et al. 2003; Wang et al. 2007). For these

reasons, nanoparticles designed for pharmaceutical appli-

cations should have extremely low size dispersities.

Nanoparticle density

Our exploratory measurements of density defined the

concentrations useful for assays in future work; when data

outside this range were eliminated, there remained a small

number of measurements allowing us to determine the

densities of the stabilised HSA, the intermediate synthesis

products, and various versions of the final nanoparticles. A

trend was evident in these densities, and although the small

data set prohibits definitive assessment, we can readily

explain the trend that we observed. Stabilised HSA was the

most dense part of the conjugate and of the final nano-

particle, with density 1.85 g/cm3 in approximate agreement

with results reported by other workers (Chick and Martin

1912). The addition of DTPA moieties reduces the density

by defining a hydrodynamic (rotator) volume or shell, in

which approximately 20 Gd-DTPA molecules protrude

from the HSA: the major part of this volume is occupied by

water and only a small fraction by DTPA. DTPA-Gd has a

density given as 1.01 g/cm3 (ch.oddb.org) or as 1.02 g/cm3

(Bayer product sheet). The larger apparent volume of the

HSA-DTPA is due mainly to water, thus reducing the

mass/volume (density). The resulting density (1.58 g/cm3)

is later raised to 1.75 g/cm3 by the addition of a dense

metal (gadolinium) in the ratio one metal ion per DTPA

moiety. The nanoparticle is constructed by packing 6–10

HSA-DTPA-Gd molecules into a volume of about

3,000 nm3, The packing density of 0.17 accounts for

essentially the entire volume of the nanoparticle, but not all

the water in the nanoparticle is bound in the HSA-DTPA-

Gd shells: as discussed in our previous paper (Stollenwerk

et al. 2010), a spheroidal volume cannot be packed 100%

with spheres, the maximum packing density being about

85%. Thus, a further *15% non-bound water is included

in the nanoparticle, in addition to the water included in the

HSA-DTPA-Gd conjugates; this reduces the nanoparticle

density yet further, to 1.36 g/cm3.

Nanoparticle integrity

The total dissolution of the particles in 1% SDS gels,

containing anionic detergents, showed that the Schiff bases

formed by glutaraldehyde crosslinking may in some cir-

cumstances be inadequate to guarantee particle cohesion.

‘‘Quenching’’ of glutaraldehyde-crosslinked nanoparticles

by application of ammonium salts resulted in nanoparticles

with a high degree of internal cohesion, and which there-

fore did not enter the gels. Ammonium salts evidently

mediate a type of strong crosslinking. The nature of this

crosslinking is unclear, but it requires the presence of both

glutaraldehyde (or of glutaraldehyde-induced Schiff bases)

and the ammonium ion. We consider it likely that nitrogen

atoms become involved in covalent bonds at the sites of the

glutaraldehyde-induced Schiff bases. The ammonium-

mediated crosslinking did not destroy HSA antigenic epi-

topes, indicating that the proteins were present in the

nanoparticles in their native conformations. Reducing the

glutaraldehyde-formed Schiff bases in the nanoparticles to

covalency by treatment with sodium cyanoborohydride

resulted in nanoparticles that were more stably crosslinked

than those treated with glutaraldehyde alone, but less stably

crosslinked than those treated with glutaraldehyde then

subsequently with ammonium ions (Fig. 4). We do not yet

know which degree of crosslinking will be appropriate

within animal tissues, so simply note that varying strengths

of crosslinking can be obtained by simple methods.

Nanoparticle relaxivity

In purified and standardised nanoparticle preparations, all

the contrast enhancement (relaxivity) is due to gadolinium

chelated to DTPA which is covalently bound to the HSA.

In series III nanoparticles, we found a relationship between

the number of gadolinium atoms on the nanoparticles and

their relaxivity, only *19% of the variability in the

relaxivity remaining unexplained (Fig. 11). In the multi-

step path leading to this data, numerous types of error can

Histochem Cell Biol (2010) 134:171–196 191

123

occur, including inaccuracies in synthesis, partial failure of

purification procedures, and measurement errors in

assessment of the concentrations of HSA, Gd, and of r1

relaxivity. In view of these numerous sources of potential

error, we consider 81% explanation of the relaxivity to

show the data set to be of good quality. It furthermore

suggests a valid comparability between the several batches

shown, indicating a good purity and standardisation level

had been achieved. Figure 11 also charts our learning

progress as we adjusted the molar concentrations of the

reagents and optimised the reaction conditions. The prin-

ciple example of this was the improvement in the Gd

loading ratio from 8 to 25 after recognising the importance

of using fresh dry DTPABA. Towards the end of this

learning process, we could reliably produce HSA-DTPA

conjugate molecules each bearing typically 19 DTPA

chelators, and responsible for [81% of the observed

relaxivity.

Description of the nanoparticles

At the nanoscale familiar scaling relationships, incorpo-

rated in our intuitive assessments of phenomena and pro-

cesses, are no longer valid (Gao et al. 2003; Wang et al.

2007). The connections between electrical charge, stability,

viscosity and many other properties are different to those

valid at the meter scale. It is therefore useful to attempt to

form an image of the nanoparticles, to facilitate under-

standing of their internal arrangement and cohesion, and of

their external interactions (Rocke 2010). The quantitative

physicochemical data obtained for this series of nanopar-

ticles was of high quality, as shown by comparable data

obtained from entirely different approaches (see for

example, Fig. 11), and was adequate to allow conceptu-

alisation of the ‘‘average’’ nanoparticle; the relatively good

standardisation of the nanoparticle properties implied that

most nanoparticles would be similar in properties to the

‘‘average’’ nanoparticle. The nanoparticle size (average

18 nm) and packing density (18–49%, depending on

whether the calculation is based on HSA only, or on HSA-

DTPA-Gd conjugates) allow the average nanoparticle to be

reconstructed as follows. The nanoparticle occupies a total

volume of 3,054 nm3, and our counts show that this vol-

ume contains 6 HSA molecules. Since one HSA molecule

occupies *90 nm3 (Stollenwerk et al. 2010; Sugio et al.

1999), the nanoparticle has packing density = 0.17.

However, this calculation ignores the presence of DTPA-

Gd chelates on the conjugate molecule, which therefore

occupies more volume than the HSA molecule alone. As

noted above, the stabilised HSA molecules have average

sizes (hydrodynamic diameters, measured in PCS) close to

10 nm, whereas the HSA-DTPA-Gd conjugate molecules

have average sizes close to 15 nm. These hydrodynamic

diameters represent freely moving (rotating, tumbling)

molecules and thus give the appearance of a solid surface.

For immobilised molecules, as found in the interiors of the

nanoparticles, the (typically) 19 DTPA-Gd chains do not

fill the surrounding volume to create a spheroidal structure.

Instead, they protrude from the HSA molecules as hairs or

filaments, and do not fill the space entirely. The DTPA

chain folds around the metal (gadolinium) ion during

chelation, resulting in a structure which is approximately

1–2 nm in diameter (Choppin and Schaab 1996; Maceke

et al. 1989). We consider here that DTPA length is

approximately 2 nm. In this case, the axes of the conju-

gate molecule are (8 ? 4) 9 (3 ? 4) 9 (3 ? 4) = 12 9

7 9 7 nm in length, so that the molecule defines a spatial

volume of *588 nm3, though its substantial mass fills only

90 nm3. On this basis, the packing density could be cal-

culated as [(6 9 588) / 3,054] 9 100% = 115%, but the

large increase in nominal packing density, from *18 to

*115%, is accounted for mainly by water. In summary,

the protein conjugate molecules occupy essentially the

entire volume of the nanoparticle, though substantially

filling only about 18% of this volume. Most of the material

in the volume defined by the presence of the DTPA-Gd

chelates is water; but by limiting the close packing of the

molecules, these chelates might define the spatial ordering

of the conjugate molecules in the interior of the nanopar-

ticle, and therefore determine the size of the nanoparticle.

Knowledge of the packing density allows us to calculate

the total weight of the average 18 nm diameter nano-

particle, as follows. Since the HSA-DTPA-Gd conju-

gate contains (on average) 19 Gd-DTPA chelates (each

with molecular weight [393.35 ? 157.25 = ] 550.6 Da),

it weighs (66,400 ? [19 9 550.6]) = 76.9 kDa. The

total weight of the average nanoparticle is there-

fore (6 9 76.9 kDa =) 0.461 MDa. We note that 1 Da

weighs 1.66 9 10-24 g so, with rounded figures, a nano-

particle of 18 nm diameter weighs 0.46 MDa, which is

7.7 9 10-19 g. It follows that the number of nanoparticles

in 1 g is 1.3 9 1018, and taking the Avagadro number of

nanoparticles to represent ‘‘1 mol’’ of nanoparticles, then

1 mol of nanoparticles weighs 461 kg, and 1 g of nano-

particles is 2.17 9 10-6 mol; a 1 g/ml solution of nano-

particles is 2.17 lM.

The PLA in the nanoparticles has been ignored in the

preceding calculations, but its concentration is so small that

it does not alter those results significantly. In the series II

particles, for which this part of the synthesis protocol was

essentially identical to series III, PLA accounted for 1–5%

of the particle mass (Stollenwerk et al. 2010). These small

amounts could not be determined with great accuracy using

the method we employed. We consider the range 1–3% in

the following. In a nanoparticle weighing 0.46 MDa, 1%

PLA would weigh *4.6 kDa, 2% PLA would weigh

192 Histochem Cell Biol (2010) 134:171–196

123

9.2 kDa and 3% PLA would weigh 13.8 kDa. One lactide

subunit (monomer) of PLA weighs 90 Da, so a PLA

polymer weighing 4.6 kDa contains 51 lactide monomers,

that weighing 9.2 kDa contains 102 lactide monomers,

and that weighing 13.8 kDa contains 153 lactide mono-

mers. The lactide monomer has length 0.5 nm, so the

three PLA chains comprising, respectively, 1, 2 or 3% of

the 18 nm diameter nanoparticles would be 25.5, 51 and

76.5 nm long. Evidently they must lie coiled within

the nanoparticle interior, because each is longer than the

nanoparticle diameter. We note that in preparing the

nanoparticles we used PLA chains nominally of 90 kDa

length, and they are much longer (1,000 monomers,

*500 nm chain length). PLA is hydrophobic, and there-

fore enters into physical and geometric configurations

which minimise its exposure to aqueous environments. In

the interior of a nanoparticle it must either associate with

HSA, or must form coils and knots to generate hydro-

phobic pockets. The six HSA conjugate molecules in our

18 nm nanoparticles would be large enough to ‘‘coat’’

about 50 nm of a PLA chain, such as the 51 nm chain

which would represent 2% of the mass in our 18 nm

nanoparticles. They would not suffice to coat the 76.5 nm

chain which would represent 3% of the nanoparticle mass,

so that a PLA chain 76.5 nm long must form coils or

knots along its length, or fold back to associate itself

repeatedly with the HSA molecules. The 6 conjugate

molecules in one nanoparticle would coat less than 10%

of the 500 nm long PLA chain (90 kDa) which we

applied when emulsifying the HSA and PLA to form the

nanoparticles. However, such long chains are not present

in our nanoparticles: we know this because, if they were

present, their mass would represent (90 kDa/461 kDa =)

19.5% of the entire nanoparticle mass. This large fraction

lies well within the detection limits of the PLA assay we

used and we would have measured it repeatedly; in fact

we always obtained PLA measurements close to 3% of

the nanoparticle mass. We therefore consider our nano-

particles to contain short PLA chains, even although our

starting material was a long PLA chain. In sum, our

evidence indicates that the nanoparticles contain chains

short enough to be fully coated by HSA, so that PLA does

not form knots in the nanoparticles. We do not know how

the short PLA chains arose, though we suspect that the

long and extremely energetic mechanical stresses of the

emulsification procedure (20,000 rpm for 15 min) might

have shredded the original long chains, by aiding hydro-

lysis of the PLA (Shih 1995; Karst and Yang 2006).

The nanoparticles contain a high proportion of water.

Since 18 g of water (= 18 cm3 at 15�C) contain 1 mol of

water, one nanoparticle of 18 nm diameter contains (3,054/

18 9 1021 =) 1.70 9 10-19 mol of water. The nanoparticle

volume thus contains (1.70 9 10-19 9 6.022 9 1023 =

102,374) *100,000 molecules of water. This calculation

ignores the *18% of the nanoparticle volume occupied by

the HSA molecules, but it should be noted that the HSA

molecule contains a significant amount of water internally.

The DTPA chelates define a volume close to the entire

volume of the nanoparticle, and we consider the water

within the ‘‘shell’’ defined by the DTPA chelates to be

bound water. Spheroidal structures cannot be packed with

packing densities higher than *0.85 (see detailed discus-

sion in Stollenwerk et al. 2010), and the HSA-DTPA-Gd

conjugates, though rather oblate, also cannot be packed

with packing densities much greater than 0.85. The water

between them, outside the shells formed by the DTPA

chelates, is not bound. This is important for fast exchange

between the unbound and bound water pools, which is a

requirement for achieving high MR relaxivity. Two fea-

tures of the nanoparticles are highlighted by considering

the water contained in the particle. First, each of the

approximately 100–200 gadolinium ions in the nanoparti-

cle is surrounded by an adequate number of water mole-

cules and this water has full access to the water

surrounding the particle, thus enabling fast exchange and

hence an increase of the relaxation rate. Secondly, the

nanoparticle should not be considered a rigid structure but

rather a gelatinous one; this consideration becomes

important in designing nanoparticle mechanical integrity.

To aid understanding and discussion, a model of the

nanoparticles as they are prepared in series III is shown in

Fig. 16.

Fig. 16 The internal structure of a ‘‘typical’’ nanoparticle, showing a

section through the centre of the nanoparticle. PLA polymer is

coloured purple, pale blue shows ‘‘internal’’ bound water of the

nanoparticle, dark grey-blue shows internal non-bound water, greyshows the HSA protein—note the heart-shaped HSA molecule at the

right side of the nanoparticle—and dark blue shows the water

surrounding the nanoparticle. A single DTPA-gadolinium chelate is

shown enlarged protruding from the right side of the nanoparticle: the

gadolinium ion is shown gold. Similar chelates protrude from the

surfaces of all HSA protein molecules in the nanoparticle, defining

shells around the HSA molecules which are filled with (bound) water

Histochem Cell Biol (2010) 134:171–196 193

123

Toxicological assays

Gd is highly toxic in non-chelated form, and recent FDA

alerts concerning its application in persons with kidney

illnesses have been well publicised in the literature (Bro-

ome et al. 2007; Kuo et al. 2007; Rinck 2008; Rofsky et al.

2008); the presence of non-chelated gadolinium on the

nanoparticles must be prevented. Arsenazo testing showed

that our nanoparticles did not contain any non-chelated

gadolinium. The results of the toxicological cell-screening

studies agreed with those carried out on the previous series

II nanoparticles (Stollenwerk et al. 2010), showing that at a

nanoparticle concentration of 2 mg/ml, no loss of viability

and no membrane damage was detected for any of the

nanoparticle batches. The particles were fully hemocom-

patible, they did not induce damage of erythrocytes

(hemolysis), activate plasmatic coagulation, nor induce

aggregation of platelets. The slightly decreased levels of

prothrombin fragments (F1 ? 2 levels) may be due to the

binding of proteins to the particle surface, a phenomenon

called formation of a ‘protein corona’ (Lundqvist et al.

2008). Only at local concentrations of 5 mg/ml could the

first evidence of toxic effects be discerned. Our investi-

gations in animal models show that concentrations

exceeding 2 mg/ml will not be necessary for imaging

studies (manuscripts in preparation).

Nanoparticles bearing PEG chains

This initial exploration of PEGylation showed that the

HSA nanoparticles were readily PEGylated with PEG

chains of sizes 5–20 kDa. PAGE analysis showed that

these chains stabilise the nanoparticles, and that this sta-

bilisation is greater for longer PEG chains (Fig. 4d). Since

stabilisation of the nanoparticles is likely to be important

for their successful targeting within living tissues, and

since variable degrees of stabilisation may be desirable,

this stabilisation by PEG chains will be an important topic

for study in subsequent work.

Nanoparticles bearing LEA targeting molecules

This initial exploration of lectin targeting showed that LEA

is readily covalently bound to HSA-based nanoparticles, by

an appropriate modification of the two-stage carbodiimide

method (Irache et al. 1994). This is generally used as a

carboxyl activating agent for the coupling of primary

amines to yield amide bonds by reaction with free amino

groups of the ligand polypeptide chains (Olde Damink

et al. 1996); in albumin, these carboxylic groups can be

found on aspartic and glutamic acid residues. We did not

determine directly the number of LEA molecules attached

to each nanoparticle, though hemagglutination studies

provided an indirect assessment of this. Hemagglutination

was used as an in vitro surrogate to test specific binding

capacity of the LEA-bearing nanoparticles to their ligands

(oligolactosamines), which in living animal tissues are

present on the surface of erythrocytes and, indeed, of all

epithelial cells (Davidson et al. 1988; Debbage 1996;

Konska et al. 2003). Hemagglutination results from the

divalent binding of LEA molecules to the erythrocytes so

that, in a blood sample containing LEA, a single LEA

molecule can bind to two different erythrocytes and cause

them to adhere to one another; many LEA molecules will

bind twice to a single cell, so the hemagglutination assay is

only semiquantitative. In contrast, in a blood sample con-

taining LEA-bearing nanoparticles, each LEA molecule

attached to one particular nanoparticle needs to bind to

only one cell to cause agglutination, because other LEA

molecules attached to the nanoparticle will bind to one or

several other cells and thus cause agglutination. Since the

orientation of the individual LEA molecules on the nano-

particles was not optimised in our work, many will be

wrongly oriented for binding their ligand: these LEA

molecules will have no agglutination effect so that, for this

reason also, the hemagglutination assay is only semi-

quantitative. The results obtained in our four nanoparticle

batches, in experiments carried out during 2 years by dif-

ferent operators, all showed that the LEA-bearing nano-

particles were comparable to LEA in their efficacy to cause

hemagglutination, and the consensus result (compare

Table 1) was that, on a molar basis, the nanoparticles were

69–129 more effective than was LEA in causing hem-

agglutination. A result like this would be obtained if each

of the 6–10 HSA-DTPA-Gd molecules packed into each

individual nanoparticle (see below) were to bear a single

LEA molecule, in correct orientation for binding its ligand

on the erythrocytes. We consider this match between data

obtained in very different fashions (hemagglutination,

TEM measurements of nanoparticles) as an indication that

our experiments reveal the physicochemical nature of the

nanoparticles accurately. Closer consideration of Fig. 4d

(lane 9), however, suggests that the LEA-bearing nano-

particles, as prepared in this work, may not have sufficient

mechanical stability to function as single units when

applied in the turbulent environment of the blood flow,

rather than in the non-turbulent environment of the test-

tube.

Upscaling of the nanoparticles

For several reasons, these nanoparticles are not yet suitable

for upscaling into industrial production. First, their sizes are

not adequately standardised; with PDI *1.2–1.6, they are

much more uniform than earlier series of the same nano-

particles (Stollenwerk et al. 2010), but industrial production

194 Histochem Cell Biol (2010) 134:171–196

123

and pharmaceutical applications would require PDI \1.1.

Some of the variability remaining in our nanoparticle bat-

ches may be due to the varying, and not fully quantified,

lengths of the PLA chains within the particles. Secondly,

the emulsification step generates nanoparticles with a cer-

tain rate of aggregation, and purification by diafiltration also

causes some aggregation, and this varies between batches.

Thirdly, the numbers of DTPA-Gd chelates attached to the

nanoparticles varies from batch to batch, and as a result the

relaxivities of the nanoparticles are not sufficiently stand-

ardised for pharmaceutical applications. Glutaraldehyde

crosslinking appears sometimes to remove a (variable)

number of DTPAs from the nanoparticles. In spite of these

blemishes, this generation of nanoparticles has several

qualities which make it interesting for application in animal

models, for example nanoparticle production in gram

amounts, and the final nanoparticles characterised by high

uniformity and high relaxivity and by enhanced stability.

The synthesis efficiencies are adequate for laboratory

preparations. Although it is not sufficiently advanced for

translational research, this generation of HSA-based nano-

particles provides excellent reagents for exploratory work

in Molecular Imaging in animal models; our next manu-

scripts in this series will explore MR imaging, biodistri-

bution and pharmacokinetics of these particles in animal

models (manuscripts in preparation). It also provides a good

starting platform for studies of drug loading.

Acknowledgments The Austrian Nano-Initiative, the Austrian

Science Foundation (FWF) (Project N201-NAN) and the Austrian

National Bank Jubilee Programme supported this work (Projects

9273, 10844, 11574 and 13096).

References

Abuchowski A, van Es T, Palczuk NC, Davis FF (1977) Alteration of

immunological properties of bovine serum albumin by covalent

attachment of polyethylene glycol. J Biol Chem 11:3578–3581

Alimarin IP, Savvin SB (1966) Application of arsenazo III and other

azo-compounds in the photometric determination of certain

elements. Pure Appl Chem 13:445–456

Andre C, Guillaume YC (2004) Zinc–human serum albumin associ-

ation: testimony of two binding sites. Talanta 63:503–508

Balbirnie M, Grothe R, Eisenberg DS (2001) An amyloid-forming

peptide from the yeast prion Sup35 reveals a dehydrated b-sheet

structure for amyloid. Proc Natl Acad Sci 98:2375–2380

Basargin NN, Ivanov VM, Kuznetsov VV, Mikhaliova AV (2000)

40 years since the discovery of the arsenazo III reagent. J Anal

Chem 55:204–210

Broome DR, Girguis MS, Baron PW, Cottrell AC, Kjellin I, Kirk GA

(2007) Gadodiamide-associated nephrogenic systemic fibrosis:

why radiologists should be concerned. Am J Roentgenol

188:586–592

Carter DC, Ho JX (1994) Structure of serum albumin. Adv Protein

Chem 45:153–203

Chen RF (1967) Removal of fatty acids from serum albumin by

charcoal treatment. J Biol Chem 242:173–181

Chick H, Martin CJ (1912) The density and solution volume of some

proteins. Chem Ind Kolloide Zeitsch 11:102–107

Choppin GR, Schaab KM (1996) Lanthanide(III) complexation with

ligands as possible contrast enhancing agents for MRI. Inorga-

nica Chim Acta 252:299–310

Davidson SE, McKenzie JL, Beard MEJ, Hart DNJ (1988) The tissue

distribution of the 3a-fucosyl-N-acetyl lactosamine determinant

recognized by the Cd15 monoclonal antibodies Cmrf-7 and 27.

Pathology 20:24–31

Debbage PL (1996) A systematic histochemical investigation in

mammals of the dense glycocalyx glycosylations common to all

cells bordering the interstitial fluid compartment of the brain.

Acta Histochem 98:9–28

Debbage P (2009) Targeted drugs and nanomedicine: present and

future. Curr Pharm Des 15:153–172

Debbage P, Jaschke W (2008) Molecular imaging with nanoparticles:

giant roles for dwarf actors. Histochem Cell Biol 130:845–875

Fehske KJ, Muller WE, Wollert U (1981) The location of drug

binding sites in human serum albumin. Biochem Pharmacol

30:687–692

Flacke S, Fischer S, Scott MJ, Fuhrhop RJ, Allen JS, McLean M,

Winter P, Sicard GA, Gaffney PJ, Wickline SA, Lanza GM

(2001) Novel MRI contrast agent for molecular imaging of

fibrin: implications for detecting vulnerable plaques. Circulation

104:1280–1285

Foster JF, Sogami M, Peterson HA, Leonard WJ (1965) The

microheterogeneity of plasma albumins. II. Preparation and

solubility properties of subfractions. J Biol Chem 240:

2503–2507

Frokjaer S, Otzen DE (2005) Protein drug stability: a formulation

challenge. Nat Rev Drug Discov 4:298–306

Gao H, Baohua J, Jager IL, Arzt E, Fratzl P (2003) Materials become

insensitive to flaws at nanoscale: lessons from nature. Proc Nat

Acad Sci USA 100:5597–5600

Goldschmidt L, Teng PK, Riek R, Eisenberg D (2010) Identifying the

amylome, proteins capable of forming amyloid-like fibrils. Proc

Natl Acad Sci 107:3487–3492

Goldwasser P, Feldman J (1997) Association of serum albumin and

mortality risk. J Clin Epidemiol 50:693–703

Gonzalez ER, Kannewurf BS (1998) Clinical review of appropriate

uses for albumin. US Pharm 23:HS15–HS26

Griffel MI, Kaufman BS (1992) Pharmacology of colloids and

crystalloids. Crit Care Clin 8:235–253

Griffiths JR, Glickson JD (2000) Monitoring pharmacokinetics of

anticancer drugs: non-invasive investigation using magnetic

resonance spectroscopy. Adv Drug Deliv Rev 41:75–89

Harisinghani MG, Barentsz J, Hahn PF, Deserno WM, Tabatabaei S,

van de Kaa CH, de la Rosette J, Weissleder R (2003)

Noninvasive detection of clinically occult lymph-node metasta-

ses in prostate cancer. N Engl J Med 348:2491–2499

Hengerer A, Grimm J (2006) Molecular magnetic resonance imaging.

Biomed Imaging Interv J 2:e8. doi:10.2349/biij.2.2.e8.

http://www.biij.org/2006/2/e8

Irache JM, Durrer C, Duchene D, Ponchel G (1994) In vitro study of

lectin–latex conjugates for specific bioadhesion. J Control

Release 31:181–188

Jaffer FA, Weissleder R (2005) Molecular imaging in the clinical

arena. JAMA 293:855–862

Karst D, Yang Y (2006) Molecular modeling study of the resistance

of PLA to hydrolysis based on the blending of PLLA and PDLA.

Polymer 47:4845–4850

Konska G, Zamorska L, Pituch-Noworolska A, Szmaciarz M, Guillot

J (2003) Application of fluorescein-labelled lectins with different

glycan-binding specificities to the studies of cellular glycocon-

jugates in human full-term placenta. Folia Histochem Cytobiol

41:155–160

Histochem Cell Biol (2010) 134:171–196 195

123

Kragh-Hansen U (1981) Molecular aspects of ligand binding to serum

albumin. Pharmacol Rev 33:17–53

Kuo PH, Kanal E, Abu-Alfa AK, Cowper SE (2007) Gadolinium-

based MR contrast agents and nephrogenic systemic fibrosis.

Radiology 242:647–649

Lundqvist M, Stigler J, Elia G, Lynch I, Cedervall T, Dawson KA

(2008) Nanoparticle size and surface properties determine the

protein corona with possible implications for biological impacts.

Proc Natl Acad Sci 105:14265–14270

Maceke HR, Riesen A, Ritter W (1989) The molecular structure of

indium-DTPA. J Nucl Med 30:1235–1239

Magnotti R (2008) Detection of gadolinium chelates. World Patent

WO 2008/045767 A2

Means GE, Feeney RE (1995) Reductive alkylation of proteins. Anal

Biochem 224:1–16

Montero EI, Benedetti BT, Mangrum JB, Oehlsen MJ, Qu Y, Farrell

NP (2007) Pre-association of polynuclear platinum anticancer

agents on a protein, human serum albumin. Implications for drug

design. Dalton Trans 43:4938–4942

Mulder WJM, Strijkers GJ, van Tilborg GAF, Griffioen AW, Nicolay

K (2006) Review article: lipid-based nanoparticles for contrast-

enhanced MRI and molecular imaging. NMR Biomed

19:142–164

Nagaraja TN, Croxen RL, Panda S, Knight RA, Keenan KA, Brown

SL, Fenstermacher JD, Ewing JR (2006) Application of arsenazo

III in the preparation and characterization of an albumin-linked,

gadolinium-based macromolecular magnetic resonance contrast

agent. J Neurosci Meth 157:238–245

Nelson R, Sawaya MR, Balbirnie M, Madsen AØ, Riekel C, Grothe

R, Eisenberg D (2005) Structure of the cross-ß spine of amyloid-

like fibrils. Nature 435:773–778

Olde Damink LH, Dijkstra PJ, van Luyn MJ, van Wachem PB,

Nieuwenhuis P, Feijen J (1996) In vitro degradation of dermal

sheep collagen cross-linked using a water-soluble carbodiimide.

Biomaterials 17:679–684

Paschkunova-Martic I, Kremser C, Mistlberger K, Shcherbakova N,

Dietrich H, Talasz H, Zou Y, Hugl B, Galanski M, Solder E,

Pfaller K, Holiner I, Buchberger W, Keppler B, Debbage P

(2005) Design, synthesis, physical and chemical characteriza-

tion, and biological interactions of lectin-targeted latex nano-

particles bearing Gd-DTPA chelates: an exploration of magnetic

resonance molecular imaging (MRMI). Histochem Cell Biol

123:283–301

Peters T Jr (1985) Serum albumin. Adv Protein Chem 37:161–245

Putnam FW (1984) The Plasma Proteins, vol 4, 2nd edn. Academic

Press, London

Rainey TG, Read CA (1994) The pharmacological approach to the

critically ill patient, 3rd edn. Williams & Wilkins, Baltimore,

pp 272–290

Rehman S, Jayson GC (2005) Molecular imaging of antiangiogenic

agents. Oncol Cancer Imaging 10:92–103

Rinck PA (2008) Radiologists meet with heavy collateral damage.

Diagnostic Imaging Europe, November 2008, pp 19–22

Robbens J, Vanparys C, Nobels I, Blust R, van Hoecke K, Janssen C,

de Schamphelaere K, Roland K, Blanchard G, Silvestre F,

Gillardin V, Kestemont P, Anthonissen R, Toussaint O, Vank-

oningsloo S, Saout C, Alfaro-Moreno E, Hoet P, Gonzalez L,

Dubruel P, Troisfontaines P (2010) Eco-, geno- and human

toxicology of bio-active nanoparticles for biomedical applica-

tions. Toxicology 296:170–181

Rocke AJ (2010) Image and reality: Kekule, Kopp, and the scientific

imagination. University of Chicago Press, USA, p 416

Rofsky NM, Sherry AD, Lenkinski RE (2008) Nephrogenic systemic

fibrosis: a chemical perspective. Radiology 247:608–612

Rohwer H, Hosten E (1997) pH dependence of the reactions of

arsenazo III with the lanthanides. Anal Chim Acta 339:271–277

Rowatt E, Williams RJP (1989) The interaction of cations with the

dye arsenazo III. Biochem J 259:295–298

Rowe JD, Bobilya DJ (2000) Albumin facilitates zinc acquisition by

endothelial cells. Proc Soc Exp Biol Med 224:178–186

Saito R, Bringas JR, McKnight TR, Wendland MF, Mamot Ch,

Drummond DC, Kirpotin DB, Park JW, Berger MS, Bankiewicz

KS (2004) Distribution of liposomes into brain and rat brain

tumor models by convection-enhanced delivery monitored with

magnetic resonance imaging. Cancer Res 64:2572–2579

Schnabel J (2010) The dark side of proteins. Nature 464:828–829

Shih C (1995) Chain-end scission in acid catalyzed hydrolysis of

poly(D, Llactide) in solution. J Control Release 34:9–15

Shrake A, Frazier D, Schwarz FP (2005) Thermal stabilization of

human albumin by medium- and short-chain n-alkyl fatty acid

anions. Biopolymers 81:235–248

Soenen SJH, Desender L, De Cuyper M (2007) Complexation of

gadolinium(III) ions on top of nanometre-sized magnetolipo-

somes. Int J Environ Anal Chem 87:783–796

Spector AA (1975) Fatty acid binding to plasma albumin. J Lipid Res

16:165–179

Stewart AJ, Blindauer CA, Berezenko S, Sleep D, Sadler PJ (2003)

Interdomain zinc site on human albumin. Proc Natl Acad Sci

100:3701–3706

Stollenwerk MM, Pashkunova-Martic I, Kremser C, Talasz H,

Thurner GC, Abdelmoez AA, Wallnofer EA, Helbok A,

Neuhauser E, Klammsteiner N, Klimaschewski L, von Guggen-

berg E, Frohlich E, Keppler B, Jaschke W, Debbage P (2010)

Albumin-based nanoparticles as Magnetic Resonance contrast

agents: I. Concept, first syntheses and characterisation. Histo-

chem Cell Biol 133:375–404. doi:10.1007/s00418-010-0676-z

Sugio S, Kashima A, Mochizuki S, Noda M, Kobayashi K (1999)

Crystal structure of human serum albumin at 2.5 A resolution.

Protein Eng 12:439–446

Suh WH, Suslick KS, Stucky GD, Suh YH (2009) Nanotechnology,

nanotoxicology, and neuroscience. Prog Neurobiol 87:133–170

Utsumi H, Yamada K, Ichikawa K, Sakai K, Kinoshita Y, Matsumoto S,

Nagai M (2006) Simultaneous molecular imaging of redox reac-

tions monitored by overhauser-enhanced MRI with 14N- and 15N-

labeled nitroxyl radicals. Proc Natl Acad Sci 103:1463–1468

Vega-Villa KR, Takemoto JK, Yanez JA, Remsberg CM, Laird

Forrest M, Davies NM (2008) Clinical toxicities of nanocarrier

systems. Adv Drug Deliv Rev 60:929–938

Veronese FM (2001) Peptide and protein PEGylation: a review of

problems and solutions. Biomaterials 22:405–417

Wang J, Karihaloo BL, Duan HL (2007) Nano-mechanics or how to

extend continuum mechanics to nano-scale. Bull Polish Acad Sci

55:133–140

Winter PM, Morawski AM, Caruthers SD, Fuhrhop RW, Zhang H,

Williams TA, Allen JS, Lacy EK, Robertson JD, Lanza GM,

Wickline SA (2003) Molecular imaging of angiogenesis in early-

stage atherosclerosis with avß3-integrin-targeted nanoparticles.

Circulation 108:2270–2274

Zalipsky S (1995a) Chemistry of polyethylene glycol conjugates with

biologically active molecules. Adv Drug Deliv Rev 16:157–182

Zalipsky S (1995b) Functionalized poly(ethy1ene glycol) for prepa-

ration of biologically relevant conjugates. Bioconjug Chem Rev

6:150–165

196 Histochem Cell Biol (2010) 134:171–196

123