Journal of Nanoparticle Research

, Volume 11, Issue 4, pp 981–988

Detection, separation, and quantification of unlabeled silica nanoparticles in biological media using sedimentation field-flow fractionation


  • Soheyl Tadjiki
    • Postnova Analytics USA
    • Department of Metallurgical Engineering, College of Mines and Earth SciencesUniversity of Utah
  • Cassandra E. Deering
    • Department of Pharmacology and ToxicologyUniversity of Utah
  • John M. Veranth
    • Department of Pharmacology and ToxicologyUniversity of Utah
  • Jan D. Miller
    • Department of Metallurgical Engineering, College of Mines and Earth SciencesUniversity of Utah
Technology and Applications

DOI: 10.1007/s11051-008-9560-3

Cite this article as:
Tadjiki, S., Assemi, S., Deering, C.E. et al. J Nanopart Res (2009) 11: 981. doi:10.1007/s11051-008-9560-3


A rapid, high-resolution methodology for characterization, separation, and quantification of unlabeled inorganic nanoparticles extracted from biological media, based on sedimentation field-flow fractionation and light scattering detection is presented. Silica nanoparticles were added to either human endothelial cell lysate or rat lung tissue homogenate and incubated. The nanoparticles were extracted by acid digestion and then separated and characterized by sedimentation field-flow fractionation. Fractions collected at the peak maxima were analyzed by transmission electron microscopy (TEM) to verify the size and shape of the isolated nanoparticles. Using the linear relationship between the particle number and the area under the fractogram, the recoveries of particles from the tissue homogenate and cell lysate were calculated as 25% and 79%, respectively. The presented methodology facilitates detection, separation, size characterization, and quantification of inorganic nanoparticles in biological samples, within one experimental run.


Sedimentation field-flow fractionationNanoparticle characterization and quantificationSilica (SiO2)Rat lung tissue homogenateHuman endothelial cell lysateTransmission electron microscopyNanocomposites


The ever-increasing production and use of nanoparticles in all aspects of technology and in the pharmaceutical industry, in particular, warrant careful studies on the effect of nanoparticles on human health. To contemplate the nanoparticle reactions in a certain environment, a thorough understanding of their physicochemical properties such as particle size distribution, dissolution, aggregation state, and surface charge/shape is required (Oberdörster et al. 2005; Borm et al. 2006). Typically nanoparticles can be detected in cells and tissues, if they have been labeled with radioactive (Nemmar et al. 2002), magnetic (Kim et al. 2006), or fluorescent markers (Kwon et al. 2008). However, the inorganic nano-sized powders being produced as bulk chemicals or the ultrafine particles produced by combustion processes do not have a convenient label that allows for their detection. As well, discrimination between nanoparticles, sub-micron particles, and soluble species with the same nominal elemental composition is a very difficult task. This complexity in detection and characterization of unlabeled nanoparticles in biological systems poses a major limitation for current research on environmental and occupational health effects of nanoparticles.

Elemental analysis methods such as Atomic Absorption Spectroscopy (AAS) or Mass Spectrometry (MS) can quantify the nonlabeled nanoparticles, but do not provide direct information on the primary particle size or aggregation state. Microscopy methods such as Transmission Electron Microscopy (TEM), offer detailed information on the size of inorganic particles down to a few nanometers (Mühlfeld et al. 2007), but obtaining an accurate statistical representation can be very tedious. Besides, sample preparation involves drying, which might result in aggregation of the sample and thus a false representation of the sample size in tissue. As well, inference of the three-dimensional structure from two-dimensional images is very difficult (Rothen-Rutishauser et al. 2007), but can be accomplished in some cases with stereological transformation (Lin and Miller 1993). Eventually, in-situ 3D characterization should be possible by nano-coupled tomography (Mühlfeld et al. 2007; Lau et al. 2008). Another common method for the size characterization of nanoparticles is dynamic light scattering (DLS) (Brant et al. 2005; Kaszuba et al. 2008), which measures a time-dependent fluctuation in the scattering intensity. A pre-defined autocorrelation function needs to be used in DLS to deduct an average diffusion coefficient for the sample. For polydispersed samples, the task is more difficult, since the signal intensity has a power relationship with size and the average size can be over-estimated if some larger particles exist in the sample. Therefore, although DLS is a common method for measuring nanoparticle size ex-situ, it is not appropriate for nanoparticle analysis in-situ or in a digested biological sample, because of the multiple scattering issues.

Field-flow fractionation (FFF) is a high-resolution method that can be an ideal candidate for direct separation, detection, and size characterization of nanoparticles in biological samples, because it allows for the separation of nanoparticles in a liquid matrix of desired composition and as a separation technique, provides a size distribution instead of an average value. Besides, the well-developed FFF theory allows calculation of the particle size from first principles and without calibration (Giddings 1988). FFF has been widely used to separate carbon nanotubes (Liu et al. 1998), particles (Giddings et al. 1994), cells, bacteria, viruses (Caldwell et al. 1981), and natural organic matter (Beckett et al. 1987). However, to the best of our knowledge it has not been utilized for characterization of nanoparticles extracted from cell or tissue samples.

Complete description of FFF theory and techniques can be found elsewhere (Giddings and Caldwell 1989; Myers 1997). Briefly, in FFF the separation is the result of the interaction between a field that is applied perpendicularly to the stream carrying sample species in an open thin channel (Fig. 1a). The applied field forces the sample species to migrate toward the lower channel wall (accumulation wall), but this migration is countered by the sample diffusion. At the beginning of a typical FFF run, the sample is injected into the channel and the flow is stopped for an appropriate “relaxation time” that allows the two motions (migration of the sample towards the accumulation wall and sample diffusion) to reach an equilibrium. During this period, particles of a given size form a cloud with thickness l (Fig. 1b). The flow profile is parabolic in the thin channel, except for the edges. This means that the flow streamlines are slower near the walls and faster near the channel center. Therefore, smaller samples with large l values, closer to the center of the channel, move with the faster stream of the flow, and elute ahead of the larger particles with smaller l value that have a stronger interaction with the field (Fig. 1b). Depending on the applied field, different information can be obtained for the sample species (e.g., nanoparticles) such as diffusion coefficient, buoyant mass, or thermal diffusion coefficient (Giddings 1988). In Sedimentation FFF (SdFFF), the channel is placed inside a centrifuge basket and is spun to create a centrifugal force as the applied field. The equivalent spherical diameter (d) of a well-retained particle can be calculated directly from the retention time (tr) (Giddings and Caldwell 1989; Ratanathanawongs Williams et al. 1997):
$$ d = \left( {\frac{{36kTt_{\text{r}} }}{{\pi wG\Updelta \rho t^{0} }}} \right)^{1/3} $$
where k is the Boltzmann constant. T is the temperature (K), w is the channel thickness (m), G is the centrifugal acceleration (m s−2), Δρ is the density difference between the particle and the carrier fluid (kg m−3). t0 is the void time (s), the time for a nonretained component to exit the channel and is given by:
$$ t^{0} = \frac{{V^{0} }}{{\dot{V}}} $$
where V0 is the channel void volume (m3) and \( \dot{V} \) is the volumetric carrier flow rate (m3 s−1).
Fig. 1

Schematic representation of separation in a field-flow fractionation channel. a The principle of FFF technique, the separation takes place in a thin, rectangular channel with triangular ends, as a result of the interaction of an incoming field applied perpendicular to the sample species. w is the channel thickness. b The cross-section of the channel. Samples are retained as a result of the interaction with the applied field and differential transportation by the laminar flow. Differences in the elevation of the particle cloud (ℓ1 and ℓ2) from the accumulation wall results in separation, since each cloud moves with a different stream of the laminar flow

Equation 1 demonstrates the resolution power of SdFFF. Under a constant field, a 2-fold increase in size should result in an 8-fold increase in retention time (tr ∝ d3). In order to reduce the operation time, it is common to decrease the applied field with time using a field decay program. The dependence of the field strength on time for power programming is given by (Williams et al. 1987):
$$ {\text{Initial}}\;{\text{rpm}} = {\text{Final}}\;{\text{rpm}}\left( {\frac{{t_{1} + 4t_{1} }}{{t + 4t_{1} }}} \right)^{4} $$
where t1 is the period of the constant field preceding the field decay.
The lower detectable size limit in sedimentation FFF depends on the particle density and the applied centrifugal force (Giddings et al. 1991):
$$ d_{\min } = \left( {\frac{36kT}{\pi wG\Updelta \rho }} \right)^{1/3} $$
For silica nanoparticles using a centrifugal force of 1,054g (2,500 rpm), the lower size limit is about 27 nm. Silica particles with sizes smaller than this limit can be characterized by other FFF techniques, such as flow field-flow fractionation (FlFFF). For inorganic particles with higher density, the minimum resolvable size can be much smaller.

Inorganic nanoparticles of interest in nanotoxicology range from amorphous silica with an approximate density of 2 g cm−3, through calcium and magnesium oxides with densities ranging from 3.3 to 3.6 g cm−3, lead and uranium oxides with densities of 9–10 g cm−3, and gold with a density of 19 g cm−3. Particles such as carbon nanotubes and C60, with densities of 1.3–1.65 g cm−3 can be separated using FlFFF, which separates the particles based on the diffusion coefficient instead of the buoyant mass.

Silica (SiO2) nanoparticles (nominal diameters 70–250 nm) were analyzed in an aqueous medium (0.1% FL-70, Fisher Scientific), to test the methodology and to replicate earlier work (Giddings et al. 1994) in aqueous systems. The same methodology was then used to detect, separate, and quantify a specific nanoparticle (d ~ 70 nm) extracted from two different biological samples, rat lung tissue homogenate and human endothelial cell culture lysate.



Silica nanoparticles with nominal diameters of 70 and 150 nm were purchased from Postnova Analytics and G. Kisker GBR (Steinfurt, Germany), respectively. The 70 nm particles were received as a suspension in deionized water and 0.01% NaN3 with a particle concentration of 25 mg mL−1. The 150 nm particles were received as a suspension in deionized water with a particle concentration of 50 mg mL−1. The 250 nm particles were received in powder form from Alfa Aesar (Ward Hill, MA). The density reported by the manufacturers for all silica particles was 2 g cm−3.

Lung tissue homogenate and cell lysate preparation

A total of 1.9 g of whole lungs from post-mortem male Sprague–Dawley rats were snap frozen in liquid nitrogen, ground with a mortar and pestle and suspended in 5.7 mL of the following mixture: 20 mM HEPES (4-(2-Hydroxyethyl) piperazine-1-ethanesulfonic acid), pH 7.9, 25% glycerol, 1.5 mM MgCl2, 0.02 M KCl, 0.2 mM ethylene diamine tetraacetic acid (EDTA), 0.2 mM phenyl methyl sulfonyl fluoride, and 0.5 mM dithiothreitol (DTT). The tissue was homogenized using a portable homogenizer (ISC BioExpress, Kaysville, UT) at 30,000 rpm and then in a motor-driven Teflon-glass homogenizer (Potter-Elvehjem, Fisher Scientific) at 900 rpm. The homogenized samples were then spiked with 2.5 mg of 70 nm silica nanoparticles.

Human aortic endothelial cells (HAEC, Cambrex, Bio Science Walkersville) were cultured in 5% CO2 at 37 °C in T-25 culture flasks in endothelial cell growth medium-2 (EGM-2, Cambrex, Bio Science Walkersville) until 90% confluent. To harvest the cells for experiments, the culture medium was removed and the cells were washed with phosphate buffer saline (PBS). Following the removal of PBS, 1 mL TrypLE enzyme (Invitrogen) was added and then removed after 1 min and the cells were incubated for 5 min. The cells were washed from the dish with fresh media and collected by centrifugation at 200g, re-suspended in 500 μl 0.1% FL-70 and were sonicated with a probe for 20 s (2 s bursts). The lysed and disrupted cell contents were then spiked with 2.5 mg of 70 nm silica nanoparticles.

Extraction of nanoparticles by acid digestion

After an hour of incubation time, the spiked tissue or cell culture sample was digested with an equal volume of 60% nitric acid solution at about 95 °C. The digested samples were then centrifuged, the pellet was washed with 1 mL of 5% nitric acid and re-suspended in 500 μL 0.1% FL-70. Samples were then layered over a saturated sucrose cushion in a test tube and were centrifuged at 21,000g for 20 min (Model 5417, Eppendorf, NY). The pellets were re-suspended in 0.1% FL-70 and sonicated with a probe for 20 s (2 s bursts). Phenol was added in a 1:1 (V/V) ratio and incubated with shaking for 5 min followed by another round of centrifugation. The pellet was then washed with ethanol and centrifuged. The final pellet was re-suspended and sonicated in 0.1% FL-70 and 0.01% sodium azide (Sigma Aldrich). For both samples the final extract volume was 0.5 mL.

Sedimentation field-flow fractionation (SdFFF)

A Postnova S101 particle fractionator, SdFFF instrument, was used to separate the silica nanoparticles isolated from the tissue and cell samples. The channel has a void volume of 3.85 mL and a thickness of 225 μm. The sample species were allowed to equilibrate under the constant field (1,800 rpm) for 8 min prior to the start of the experiment. After an additional 6 min (t1), a power decay function was used to bring the field to 200 rpm at the end of the experiment. The sample injection volume was 1.7–100 μL depending on the sample. UV absorbance at 254 nm (Postnova PN3211) and light scattering (BI-MWA, Brookhaven Instruments, NY) at the 90° angle, with a laser power of 35 mW at a wavelength of 635 nm were used for sample detection. All the experiments were carried out at 295 K. 0.1% FL-70 at pH ~ 9.5 with a flow rate of 2 mL min−1 was used as the carrier fluid.

Results and discussion

Methodology verification

Fractograms of three different nanosilica particles with nominal diameters of 70, 150, and 250 nm in 0.1% FL-70, are depicted in Fig. 2. The mixture was separated using the power decay program as explained in the “Experimental” section. Since the light scattering signal strongly depends on size, a lower injection mass was used for larger particles. The small peak at about 2.5 min is the “void peak” representing the material not retained by the field. Particle sizes calculated from FFF theory (Eq. 1) are given on the upper axis. For the three individual particles and for the mixture, the size obtained at the peak maxima corresponds very well to the nominal particle sizes. The fractograms show a symmetric distribution, indicating that under the conditions of our experiments no significant aggregation of nanoparticles had taken place. The fractogram of the mixture (black line) shows excellent resolution between the distributions of individual samples.
Fig. 2

Fractograms showing the separation of three nanosilica standards (nominal diameters 70, 150, and 250 nm). The gray lines show the fractograms of the particles injected separately and the solid black line shows the fractogram of the mixture. The upper axis shows the corresponding size, calculated from Eq. 1. The small peak at about 2.5 min is the “void peak” representing the material not retained by the field

Characterization of nanoparticles isolated from tissue homogenate and cell lysate

Figure 3 shows the fractograms of 70 nm silica particles extracted from lysed human lung cell and rat lung tissue homogenate, respectively. The normalized size distributions (Fig. 4) were constructed from the light scattering detector response and SdFFF theory (Eq. 1), using Postnova FFF Analysis program. The fractograms were corrected for the baseline, and the void peaks were removed prior to construction of size distributions.
Fig. 3

Fractograms of the silica nanoparticles extracted from lung tissue homogenate (gray line) and lung cell lysate (black line)
Fig. 4

Normalized size distributions of silica nanoparticles extracted from a lung tissue homogenate and b lung cell lysate, constructed from the light scattering detector response and SdFFF theory (Eq. 1), using Postnova FFF analysis program. The fractograms were corrected for the baseline and the void peaks were removed prior to construction of size distributions. The inserted pictures are the TEM micrographs of the fraction collected from the peak maximum. For both samples excellent agreement is observed between the nominal size for the silica nanoparticles and the diameter at the peak maxima, calculated from the SdFFF theory (Eq. 1)

To verify the size and shape of the separated nanoparticles, and to confirm that the size distribution represents the silica nanoparticles and not the components from the cell or tissue or the growth medium, fractions were collected from the peak maxima of both size distributions and were analyzed by a Philips 420 TEM, operating at 80 KV. The micrographs show individual, mono-dispersed silica nanoparticles indicating that SdFFF has been successful in separation of silica nanoparticles and that the size of the particles are comparable with the values calculated from FFF theory (Eq. 1).

Quantification of the isolated nanoparticles

In the case of optimum separation and absence of aggregation by concentration, the area under the fractograms should have a linear relationship with the number of injected nanoparticles, which can be used as a calibration curve for quantification of extracted nanoparticles. Different masses of the 70 nm nanosilica particles were injected (Fig. 5) and the area under the fractogram curves, obtained by UV and light scattering detectors was measured. The inset in Fig. 5 shows an excellent correlation between the area under the fractogram (detected by light scattering) and the number of particles, and can be used as a calibration curve. The instrumental detection limit for the light scattering detector, obtained by multiple blank runs, corresponded to 1.6 μg or 3.8 × 109 particles.
Fig. 5

Fractograms of silica nanoparticles with a nominal diameter of 70 nm, injected at different concentrations. Number of particles in each fractogram is shown next to the pointing arrows. The inset shows the calibration curve obtained using the particle number versus fractogram peak area

Using the calibration in Fig. 5, the total number of particles (particles in 0.5 mL final digest volume) were calculated as 1.75 × 1012 (625 μg) for the tissue extract and 5.5 × 1012 (1,964 μg) for the cell extract, corresponding to 25% and 79% of the initially added particles (2.5 mg), respectively.

Typical nanoparticle concentrations used for in vitro cell culture exposures are 10–300 μg mL−1 and a usual deposited dose in animal exposures is 1–5 mg particles per kg of body weight (Warheit et al. 2006; Veranth et al. 2007). A study using intravenously administered titanium dioxide nanoparticles reported deposited burdens of 2–130 μg g−1 in target organs (Fabian et al. 2007). Intratracheal instillation of TiO2 nanoparticles ranging from 30 to 2,000 μg to rats have also been reported (Oberdörster 2000). For this preliminary investigation, we used a high initial number of particles. However, considering the low detection limit (1.6 μg or 3.8 × 109 particles), the SdFFF technique demonstrated in this article is capable of detecting unlabeled inorganic nanoparticles at the concentrations observed in toxicology studies using superphysiological doses.

It should be noted that the presented methodology aims for detection and characterization of primary nanoparticles in biological matrices. Therefore, a deliberate attempt was made to fully disperse the silica nanoparticles. The same methodology can be used for detection of nanoparticle aggregates as well and the separation can be verified by independent imaging techniques as has been done in this study. For example Barman and Giddings (1992) used SdFFF for separation and characterization of polymethylmethacrylate (PMMA) particles in singlet, doublet, and triplet states. Also Moon and Giddings (1992) used SdFFF for separation of 300 nm gold particles from their aggregates. In both studies, the separation was confirmed using electron microscopy. In the case of a sample with unknown density the particle retention can be measured in carrier fluids with different densities and both particle mass and density can be extrapolated (Giddings et al. 1981; Yonker et al. 1985). A more complicated condition would be a mixture of particles with varying densities and aggregation states, in which case an auxiliary detection method such as Inductively-Coupled Plasma Mass Spectroscopy can be used (Taylor and Garbarino 1992; von der Kammer et al. 2004).

A concurrent study by the authors using tissue digestion with protease enzymes showed comparable particle detection sensitivity (data not shown here). Enzyme digestion would be more appropriate for acid-soluble particles. Further research involves improvement of the detection limit and quantification of nanoparticles in polydispersed and multi-component systems.


A rapid, high-resolution methodology based on sedimentation field-flow fractionation and light scattering detection is presented for characterization and quantification of nanoparticles extracted from tissue homogenate and cell lysates. Fractograms of a mixture of silica nanoparticles with nominal diameters of 70, 150, and 250 nm, illustrate the capability of the current SdFFF methodology for detection of nanoparticles in a single run with high resolution. The 70 nm particles were isolated from tissue homogenate and cell culture lysate, then separated and characterized by SdFFF, and the size and shape of the isolated particles were verified by examining fractions collected from the peak maxima using TEM. The linear relationship between the particle number and the area under the fractograms was used to quantify the isolated nanoparticles. Using multiple blank injections, a detection limit of 1.6 μg or 3.8 × 109 particles was calculated for the system used.

The presented methodology facilitates detection, size characterization, and quantification of nanoparticles, within one SdFFF run. This is almost impossible to achieve by other methods (microscopy or elemental analysis), which result in lost information regarding either quantity or size distribution. The presented methodology can be easily used for separation, detection, and characterization of unlabeled inorganic nanoparticles in complex biological matrices. Further research will include additional improvement of the detection limit, using more sensitive detectors such as online ICP-MS and developing calibration methods for multi-component or polydispersed systems.


The authors would like to thank Professor Marcus N. Myers for very helpful comments on the draft and Ms. Nancy Chandler at the HSC Core Research Facilities, University of Utah, for help with TEM images of the isolated nanoparticles. The authors are also grateful to Ms. Dorrie Spurlock for proofreading the manuscript.

Copyright information

© Springer Science+Business Media B.V. 2008