Journal of Nanoparticle Research

, Volume 12, Issue 1, pp 47–53

Characterisation of nanoparticle size and state prior to nanotoxicological studies


  • Iker Montes-Burgos
    • Centre for BioNano Interactions, School of Chemistry & Chemical BiologyUniversity College Dublin
  • Dorota Walczyk
    • Centre for BioNano Interactions, School of Chemistry & Chemical BiologyUniversity College Dublin
    • NanoSight Ltd.
  • Jonathan Smith
    • NanoSight Ltd.
  • Iseult Lynch
    • Centre for BioNano Interactions, School of Chemistry & Chemical BiologyUniversity College Dublin
  • Kenneth Dawson
    • Centre for BioNano Interactions, School of Chemistry & Chemical BiologyUniversity College Dublin
Special focus: Safety of Nanoparticles

DOI: 10.1007/s11051-009-9774-z

Cite this article as:
Montes-Burgos, I., Walczyk, D., Hole, P. et al. J Nanopart Res (2010) 12: 47. doi:10.1007/s11051-009-9774-z


Before commencing any nanotoxicological study, it is imperative to know the state of the nanoparticles to be used and in particular their size and size distribution in the appropriate test media is particularly important. Particles satisfying standards can be commercially purchased; however, these invariably cannot be used directly and need to be dispersed into the relevant biological media. Often such changes in the environment or ionic strength, or a change in the particle concentration, results in some aggregation or a shift in the particle size distribution. Such unexpected aggregation, dissolution or plating out, if unaccounted for, can have a significant effect on the available nanoparticle dose and on interpretation of any results obtained thereafter. Here, we demonstrate the application of characterisation instrumentation that sizes nanoparticles based on their Brownian motion in suspension. Unlike classical light-scattering techniques, the nanoparticle tracking and analysis (NTA) technique allows nanoparticles to be sized in suspension on a particle-by-particle basis allowing higher resolution and therefore better understanding of aggregation than ensemble methods (such as dynamic light scattering (DLS) and differential centrifugation sedimentation (DCS)). Results will be presented from gold (standard) nanoparticles in biologically relevant media that emphasise the importance of characterisation of the nanoparticle dispersion. It will be shown how the NTA technique can be extended to multi-parameter analysis, allowing for characterization of particle size and light scattering intensity on an individual basis. This multi-parameter measurement capability allows sub-populations of nanoparticles with varying characteristics to be resolved in a complex mixture. Changes in one or more of such properties can be followed both in real time and in situ.


Protein coronaNanoparticlesDispersionNTADLSNanoparticle tracking and analysisEnvironmentEHS


The aim of this articleis to highlight the need to measure the size distribution of nanoparticles that are to be used in nanotoxicological studies, not only in water or buffer, but also in the actual test medium that will be used for the studies. Size and size distribution are considered to be among the key parameters in determining the interaction of nanoparticles with living systems. The Royal Society of Chemistry suggested that 100 nm is the cut-off above which nanoparticles will not enter cells via receptor-mediated processes (Royal Society of Chemistry and Royal Academy of Engineering 2005), and some experimental evidence corroborating this size as a rough guide is emerging (Chithrani and Chan 2007; Clift et al. 2008). Other important size cut-offs are that particles less than 40 nm can enter the nucleus, while particles less than 35 nm can potentially cross the protective epithelial barriers, such as the blood–brain barrier (Oberdorster et al. 2004). One should be aware that the real size “cut-offs” are dependent on the material and surface details, and these values are at best only guidelines.

While many researchers check the nanoparticle manufacturers’ sizing specification, it is still uncommon for many researchers to fully consider the implications of other modifications that they make to the suspension. Changes in ionic strength and charge screening, or binding of proteins and other biomolecules to the nanoparticles surface can alter their stability in dispersion, leading to partial aggregation, and altered (unknown) concentration of dispersed nanoparticles. Aggregated particles (where the aggregates are long-lived) are no longer available for uptake by cells, and as such, meaningful exposure doses cannot be determined, making dose–response curves unreliable. In addition, many of the commercially available particles differ significantly in terms of their physical properties compared to those specified by the manufacturers (Lundqvist et al. 2008). Thus, poorly characterised samples have the potential to lead to, at best, confusing and, at worst, mis-interpreted results.

Assessing the potential biological impacts of nanomaterials has become of enormous importance in recent years, as the rapid pace of development of nanotechnology has not been matched by a complete investigation of their safety. The same properties that make nanoparticles exciting for applications, namely their small size, their enormous surface area and their high reactivity, also make them accessible to previously inaccessible locations in living systems with potentially significant consequences for nanomedicine and nanosafety. The large surface area means that they bind proteins and other biomolecules from biological solutions with great efficiency, and with much higher specificity than flat surfaces of equivalent materials (Cedervall et al. 2007a, b). Protein adsorption occurs immediately upon contact with biological media, such as tissue culture media which is often supplemented with 10% foetal calf serum, which results in altered size and size distribution of the dispersion. Thus, characterisation of the nanoparticle dispersion in the relevant test media is crucial in order to understand the nature of the dispersion actually being presented to the cells, tissue, or organism.

Here, we highlight the effect of changing the nanoparticle dispersion environment on the dispersion characteristics—size and size distribution, by addition of human plasma to represent a biological fluid. The effect of plasma on an extremely well characterised (size standard) gold nanoparticle of 60 nm is investigated. The use of a recently developed technology, NTA, for characterising such dispersions is highlighted and discussed, and the results are compared to more conventional approaches such as DLS.

Sizing techniques

Two different techniques were used to measure the dispersion characteristics—size and size distribution: dynamic light scattering (DLS) (also known as photon correlation spectroscopy (PCS) and nanoparticle tracking and analysis (NTA). In particular, the most recently developed system, NTA, was assessed in-depth due to its ability to see and size particles individually on a particle-by-particle basis (DLS being an ensemble technique where the measurement of an ensemble of particles is used to calculate a particle size distribution). The two approaches have different strengths, and are, in many circumstances, complementary.

NTA allows individual nanoparticles in a suspension to be microscopically visualized (though not, of course, imaged) and their Brownian motion to be separately but simultaneously analysed and from which the particle size distribution (and changes therein) can be obtained on a particle-by-particle basis. This enables separation of particle populations by size and intensity (Fig. 1) and allows complex and heterogeneous samples to be fully characterised without the intensity bias towards large particles which limits the usefulness of other light scattering techniques, and can result in a small number of larger particles or aggregates masking the presence of large numbers of nanoscale particles. This has huge implications for regulation of micro- and nano-materials, as it is known by colloid and particle scientists that even micron-sized materials can often have a significant tail of their size distribution in the nanometre range. As this tail may not be detected using the industry standard methods (including DLS), it is unlikely to be reported or even fully understood by many of the industry users of these materials.
Fig. 1

The resolution of two particle populations (90 and 130 nm latex) resolved here by size and intensity

In practice, the NTA technique requires a small (250 μL) sample of liquid containing particles at a concentration in the range 107–109/mL to be introduced into a scattering cell through which a focused laser beam (approx. 40 mW at λ = 635 nm) is passed (Malloy and Carr 2006). Particles within the path of the specially configured beam are observed via a microscope-based system or a dedicated non-microscope optical instrument (Fig. 2) (NanoSight LM10 and LM20, respectively) onto which is fitted a CCD camera. The motion of the particles in the field of view (approx. 100 × 100 μm) is recorded (at 30 fps) and the subsequent video analysed. Each and every particle visible in the image is individually but simultaneously tracked from frame-to-frame, and the mean square displacement is determined by the analytical program from which the particle’s diffusion coefficient can be obtained. Results are displayed as an equivalent hydrodynamic diameter particle distribution, calculated from the measured diffusion coefficient. The only information required to be input is the temperature of the liquid under analysis and the viscosity (at that temperature) of the solvent in which the nanoparticles are suspended. Otherwise, the technique is one of the few analytical techniques which is absolute and therefore requires no calibration. Notably, because the instrument visualizes particles on an individual basis, particle number concentration is recoverable. Once analysed, the sample is simply withdrawn from the unit for re-use, if required.
Fig. 2

Schematic of NTA apparatus set-up

While results can be obtained in typically 30–60 s, in this study much longer (166 s) videos were captured. In addition, the instrument was programmed to carry out repeat measurements (batches).

Methods and materials

Blood was taken from seemingly healthy donors. The blood donation was arranged such that the blood samples were labelled anonymously. They could not be traced back to a specific donor; however, it was possible to use plasma from just one of the donors for a specific experiment. The tubes were centrifuged for 5 min at 800 RCF to pellet the red and white blood cells. The supernatant (the plasma) was transferred to labelled tubes and stored at −80°C until used. Upon thawing, the plasma was centrifuged again for 3 min at 16.1 kRCF to further reduce the presence of red and white blood cells. The supernatant was transferred to a new vessel taking care not to disturb the pellet.

NIST gold standard nanoparticles of 60 nm were used (NIST reference material 8013). These were stored, prepared and used according to the relevant reports of investigation (National Institute of Standards & Technology 2007). The gold was diluted to a concentration of approximately 108 particles/mL using standard citrate buffer of pH = 7.19. For the dispersions in plasma, the human plasma was diluted 1:1,000,000 in citrate buffer, and 10 µl of gold nanoparticles were diluted with 790 µl of the diluted plasma.

Nanoparticle tracking and analysis was carried out on a NanoSight LM10 (NanoSight Ltd., UK). All sample preparation and measurements were carried out at UCD and analysis was performed by NanoSight using a beta version of NTA 2.0 software. Multiple videos, each 166 s long, were recorded and analysed in batch mode to ensure statistical invariance. For the measurements of particle size in diluted plasma, it was possible to gate out protein aggregates due to their low intensity of scattering when compared to gold nanoparticles. However, this issue of protein aggregation, given that the plasma is a natural ionic medium, will always be problematic. This question needs to be addressed further in the future.

Two depictions of the NTA results are displayed below; an uncorrected and a corrected result. The NTA technique, as described above, tracks particles for a finite number of steps and calculates a size based on the average distance travelled. This uncorrected result therefore results in an inherent breadth to the distribution. This can be readily understood by considering a particle, sized after tracking over 10 frames. As Brownian motion is a random effect, it is possible that the particle has moved either further or less far than would be predicted solely by the Stokes–Einstein equation. If the particle has moved further/faster than would be predicted by the Stokes–Einstein equation, applied to a particle with its size, then the size calculated by the software will be smaller than the particle’s true size and vice versa. Repeating this for many particles of the same actual size will build up a measured particle size distribution with an inherent breadth greater than that of the actual particle size distribution (Saveyn et al. 2008). This breadth of the distribution is fully modellable, and this model is implemented in the latest version of the software NTA 2.0, to take into account the inherent breadth of a distribution. Based on this, it is possible to infer the true particle size distribution for monomodal and bimodal samples with varying degrees of polydispersity.

Comparative experiments were performed using DLS (Zetasizer Nano ZS ZEN3600, Malvern, UK). For DLS experiments, 2 μl of gold stock solution was diluted with 2 mL of citrate buffer or citrate-diluted plasma. From this measurement, the mean size of particles inside the sample is obtained along with the correlation between the number of particles of a particular size versus the size of the nanoparticles.


Size and size distribution of the 60 nm NIST gold nanoparticles in the absence and presence of human plasma are shown in Fig. 3 as measured directly by the NanoSight method. These are the raw data in the absence of consideration of the inherent randomness of Brownian motion. The data shows the average of 15 measurements for the pure gold nanoparticles, and the average of 7 measurements for the nanoparticles dispersed in plasma. Each of the results is normalized to the peak concentration and the error bars represent the deviation between the batches. Each individual measurement comprises results from ~4,000 tracks of particles. This does not mean that 4,000 different particles were measured; as the same particles may be measured multiple times during the 166 s of data collection.
Fig. 3

Uncorrected normalized particle size distributions of 60 nm NIST gold nanoparticles (solid line) and in presence of plasma (dashed line)

The NIST gold particles are a certified size standard particle, and as such are expected to have a single peak and a relatively narrow distribution. The observed peak (in the absence of plasma) centres at 55 nm, in reasonable agreement with the manufacturer’s stated size. The observed breadth of the size distribution is about 30 nm each side of the peak which is wider that would be expected for such particle size standards. Addition of plasma resulted in a small increase in the peak position, but not much change in the peak distribution, resulting in a size of 64 ± 30 nm.

Following application of the modelling described above, the peak distribution for the 60 nm NIST gold nanoparticles in the absence of protein becomes much narrower, as expected. The size recorded is 60 ± 6 nm, in excellent agreement with the manufacturer’s size (Fig. 4) and specifications (National Institute of Standards & Technology 2007).
Fig. 4

Corrected particle size distribution of 60 nm NIST gold nanoparticles (solid line) and in presence of plasma (dashed line)

Treating the uncorrected data resulting from the presence of plasma in a similar manner, results in the peak of the nanoparticle–protein complex centering (having a modal value) at 70 nm, and the distribution is slightly wider than in the absence of plasma. This is not unexpected, as the presence of proteins in the nanoparticle dispersion will have multiple effects on the dispersion, including potentially screening of the surface charges that help to maintain the repulsion between the particles, or depletion or bridging type interactions. In addition, either due to different proteins binding or due to variation in the surface chemistry of the gold nanoparticles, it would be expected that there would be an increase in variation of the size of nanoparticles, leading to a broader particle size distribution.

The results from the two corrected graphs are summarized below along with their errors (Table 1). For each distribution, there is a mean (the average particle size measured), a mode (the most frequent particle size found, post model correction) and the standard deviation (the breadth of the fitted log-normal distribution fitted to the data). In addition, for each value there are errors given; this is the standard deviation of each of the measured parameters.
Table 1

Summary of NTA results showing corrected results for the NIST gold both in the absence and presence of plasma



Value (nm)

Error (nm)

60 nm gold (no plasma)







Standard deviation



60 nm gold (plasma)







Standard deviation



The measurements of gold in the presence and absence of plasma were also made by DLS and the results from this are shown below in Fig. 5.
Fig. 5

Normalised DLS measurements of 60 nm NIST gold nanoparticles (solid line) and in presence of plasma (dashed line)

The DLS results show some similarities to those measured by NTA. The modal sizes without and with plasma present are relatively similar at 58.5 and 70.1 nm, respectively. It is also clear that the breadth of the size distribution increases significantly with plasma present compared to without, as demonstrated by NTA.

However, there are also notable differences in the DLS and NTA results as, in both measurements, DLS shows a bimodal distribution with separate smaller peaks at 14.1 and 12.4 nm, respectively. The origin of these peaks is hard to understand and is likely to be an artifact of the DLS technique. Whilst in the presence of plasma a smaller population may be expected due to the presence of proteins and protein aggregates, the presence of such a peak in the absence of plasma makes the result confusing and hard to interpret. While the peak is significantly lower (approximately 8% of the primary mode), the size is smaller. Converting this from the plotted intensity graph to a number distribution shows a stark change in relative significance. Assuming Rayleigh scattering (suitable at these dimensions), this would relate to a number equivalent of (60nm/13 nm)6 or ~10,000, implying (if material properties were the same) approximately a 1,000× greater number of the smaller particles than of the known gold nanoparticles. Given the absence of this peak in the plasma solution alone, and the dilution of the plasma, we do not consider this a real peak.

The size of the NIST gold nanoparticles has also been measured as part of the validation process for a standard material using several other techniques, and this data is summarized below in Table 2. As can be seen the batch was (by most methods) measured to be around 55 nm.
Table 2

Average sizes as determined by various techniques and including 95% uncertainty levels (Hackley 2008)


Analyte form

Nominal 60 nm


Dry, deposited on substrate

55.4 ± 0.3


Dry, deposited on substrate

54.9 ± 0.4


Dry, deposited on substrate

56.0 ± 0.5


Dry, aerosol

56.3 ± 1.5

DLS (173°)

Diluted liquid suspension

56.6 ± 1.4

DLS (90°)

Diluted liquid suspension

55.3 ± 8.3


Native liquid suspension

53.2 ± 5.3

The change in size recorded in Fig. 4 (following the modelling) shows an increase in the nanoparticle size of approximately 10 nm, which corresponds to an approximately 5 nm thick adsorbed protein layer covering the nanoparticle, as shown schematically in Fig. 6. Assuming an approximate size of 5 nm for a typical protein (e.g. human serum albumin has dimensions of 3 × 8 × 8 nm, and is one of the larger proteins in plasms), this suggests that depending on type and orientation of the proteins bound, the protein layer surrounding the gold nanoparticles may be a monolayer or composed of several protein layers.
Fig. 6

Schematic representation of the increase in diameter of nanoparticles upon dispersion in plasma. Protein and particle not to scale. Typically several different proteins constitute the nanoparticle protein corona


This article has identified that there is a significant and measurable change in particle size distribution of monodisperse gold nanoparticles in the presence of a biological medium such as human plasma. It has identified a suitable methodology for the measurement of the thickness of the plasma layer using two independent methods, NTA and DLS, which have been used to validate one another.

The breadth of the distribution as measured by both DLS and NTA is similar to that determined using DLS in the standard particle characterisation by NIST, and both measurements show a relatively small breadth of distribution for the peak centered around 60 nm. In the presence of plasma, both the particle size and the particle size distribution increase slightly, as expected. Further work is required to investigate the effect of plasma concentration on the thickness of protein layer formed as early indications are that this may have a very significant effect.

In addition to the measurement of particle size and size distribution, NTA also offers the ability to measure absolutely the concentration, and it is anticipated that this can be used to also give a number concentration of the NIST gold nanoparticles. NTA is also ideal for identifying and counting nanoparticle aggregates such as dimers due to its ability to visualize the nanoparticles individually. It thus represents (at least for the nanoparticles and concentration regimes where we have worked) a new, useful tool in the armoury of techniques for bionanoscience and bionanointeractions on engineered nanomaterials in biology.


Funding is gratefully acknowledged from EU FP6 project NanoInteract, NMP4-CT-2006-033231 (IL, KD), and SFI Research Frontiers Projects, PHY0033 and CHP031 (DW, IM-B).

Copyright information

© Springer Science+Business Media B.V. 2009