Methods xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Methods

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

Gold nanoparticles explore cells: Cellular uptake and theiruse as intracellular probes

http://dx.doi.org/10.1016/j.ymeth.2014.02.0061046-2023/� 2014 Elsevier Inc. All rights reserved.

⇑ Corresponding author at: Institute of Life Sciences and Department ofChemistry, University of Southampton, Highfield Campus, SO17 1BJ Southampton,UK.

E-mail address: s.mahajan@soton.ac.uk (S. Mahajan).

Please cite this article in press as: A. Huefner et al., Methods (2014), http://dx.doi.org/10.1016/j.ymeth.2014.02.006

Anna Huefner a,e, Dedy Septiadi a,b, Bodo D. Wilts a, Imran I. Patel a,c,e, Wei-Li Kuan d,Alexandra Fragniere d, Roger A. Barker d, Sumeet Mahajan a,e,⇑a Sector for Biological and Soft Systems, Cavendish Laboratory, Department of Physics, University of Cambridge, 19 JJ Thomson Avenue, Cambridge CB3 0HE, UKb Institut de Science de d’Ingénierie Supremoléculaires, Université de Strasbourg, 8 Allée Gaspard Monge, 67083 Strasbourg Cedex, Francec Cancer Research UK Cambridge Institute, University of Cambridge, Li Ka Shing Centre, Robinson Way, Cambridge CB2 0RE, UKd John van Geest Centre for Brain Repair, University of Cambridge, Forvie Site, Robinson Way, Cambridge CB2 0PY, UKe Institute of Life Sciences and Department of Chemistry, University of Southampton, Highfield Campus, SO17 1BJ Southampton, UK

a r t i c l e i n f o a b s t r a c t

Article history:Received 22 October 2013Revised 24 January 2014Accepted 6 February 2014Available online xxxx

Keywords:Gold nanoparticlesNanoparticle uptakeIntracellular imagingSurface-enhanced Raman spectroscopySERS

Understanding uptake of nanomaterials by cells and their use for intracellular sensing is important forstudying their interaction and toxicology as well as for obtaining new biological insight. Here, we inves-tigate cellular uptake and intracellular dynamics of gold nanoparticles and demonstrate their use inreporting chemical information from the endocytotic pathway and cytoplasm. The intracellular goldnanoparticles serve as probes for surface-enhanced Raman spectroscopy (SERS) allowing for biochemicalcharacterisation of their local environment. In particular, in this work we compare intracellular SERSusing non-functionalised and functionalised nanoparticles in their ability to segregate different but clo-sely related cell phenotypes. The results indicate that functionalised gold nanoparticles are more efficientin distinguishing between different types of cells. Our studies pave the way for understanding the uptakeof gold nanoparticles and their utilisation for SERS to give rise to a greater biochemical understanding incell-based therapies.

� 2014 Elsevier Inc. All rights reserved.

1. Introduction

Stem cells, understanding molecular mechanisms underlyingthe development of cancer and neurodegenerative diseases as wellas studying the interactions of pathogens such as viruses andbacteria with cells are topical areas in molecular biomedicalresearch. Though fluorescence microscopy is a major tool for cellu-lar imaging, it has known limitations such as the invasiveness ofthe labelling (or staining) process, inability to monitor multiplemolecular targets interactions as well as it suffers from photoble-aching. Therefore, new imaging techniques and approaches thatcan identify molecules and interactions at sub-cellular resolutionare very much required. One of the ways to achieve these objec-tives is to combine the use of nanoparticle-probes with intracellu-lar molecular spectroscopy. Gold nanoparticles (AuNPs) can beused as probes and they have recently attracted a lot of attentionin biomedical research due to their chemically inert nature and

remarkable optical properties. AuNPs have been utilised in manyapplications such as biochemical sensing [1], drug and gene deliv-ery [2,3]. The rich optical properties of AuNPs arise from their abil-ity to localise surface plasmons. This also enables their use astransducers for surface-enhanced Raman spectroscopy (SERS).

SERS is a highly sensitive, label-free and non-destructive meth-od which allows for molecular identification giving it severaladvantages over other imaging techniques such as fluorescence,infrared, UV–visible or NMR [4–7]. In SERS, the otherwise extre-mely weak Raman scattering signal [8] is enhanced by several or-ders of magnitude by ensuring that the molecule is in very closevicinity of nanoparticles or of a nanostructured gold or silver sur-face [7,9,10]. The SERS spectrum is a vibrational ‘fingerprint’ whichcharacterises the chemical bonds and symmetry of the molecule.[4] Recently, nanoparticle-based SERS has been employed in manybiomedical applications from chemical sensing [11–13] to cancerdetection in body fluids and tissues [14–17].

In contrast to conventional methods stated earlier, higherdetection limits in addition to a complete structural characterisa-tion of the target molecule can be achieved with SERS [18] withoutthe need for staining or expressing fluorogenic proteins. It has alsobeen demonstrated that SERS can be applied to living cells in order

2 A. Huefner et al. / Methods xxx (2014) xxx–xxx

to monitor cellular functions [19], cell response to stress [20] andapoptosis [21]. SERS has been extensively applied in biomedicalapplications often for label-free detection of biomolecules in cellsand tissues [4,22]. AuNPs have been used as intracellular probesfor SERS [23] to monitor release of drugs inside cells [24] as wellas to probe molecules by targeting them to cellular compartmentssuch as mitochondria, endosomes and the cell nucleus[11,22,25,26]. The key for utilisation of nanoparticles inside cellsfor SERS and other measurements is their mechanism of uptake.Particle uptake has been achieved using physical methods suchas microinjection [27,28], electroporation [29], sonoporation[23,30] and gene gun [31]. Further options for involuntary uptakeare the translocation of particles through the plasma membranefacilitated by cell penetrating peptides (CPPs) [32,33] such as thetrans-activating transcriptional protein derived from viruses [34].

Moreover, AuNPs are known to be taken up by cells voluntarilywhereby their intracellular uptake and distribution depends onmany factors such as particle characteristics, e.g. charge, surfacemodification, particle size and shape [35–39] as well as experimen-tal procedures involving concentration and exposure time [40].Different cellular mechanisms are involved in particle uptake suchas phagocytosis and pinocytosis. The latter facilitates the uptakeprocess through small membrane-bound vesicles, called endo-somes.[41,42] Pinocytosis can involve energy-dependent andreceptor mediated endocytosis which has been shown to be thedominant internalisation pathway for several cell lines[36,40,43]. Following internalisation, membrane-bound vesiclesencapsulating the particles mature and eventually fuse with lyso-somes [41].

Hence, intracellular AuNPs are trapped inside membrane-boundvesicles of the endocytotic pathway [37]. Escape from these vesiclescan only be achieved by functionalisation of the particle surfacewith peptides such as CPP [33,34] and adenoviral receptor-medi-ated endocytosis peptides [44]. Trapped in endocytotic vesicles,particles get transported through cells via the common cellulartransport mechanisms: by molecular motors such as myosin,kinesin and dynein along the intracellular filament network [45].Studies have revealed various characteristics of these motor pro-teins in vitro and in vivo with reference to endocytotic organellesbeing transported. Friedman and Vale measured the in vivo mobilityof kinesin unattached to a surface using single molecule assays tobe 600–800 nm/s [46–48]. Further studies revealed that averagevelocities strongly depend on the size of the attached cargo thatis transported by these motors: a larger size appears to correlatewith slower motions [49]. For example, 30 nm quantum-dotstagged to kinesin showed an average velocity of 600 nm/s in HeLacells [49], whereas 1 lm collagen-coated beads in murine embry-onic fibroblasts displayed a velocity of �10 nm/s [50] and 3 lmpolystyrene beads in SV80 human fibroblasts showed similar veloc-ities of 8–30 nm/s [51] suggesting that many factors influence thetransport speed in cells. These factors include variations withinthe cell line, the bead size and its surface properties and materialcomposition, which change the stall force of the motors and sterichindrance within the cell [50]. Moreover, the velocity of endocy-totic vesicles such as endosomes in budding yeast and lysosomesin African green monkey kidney cells was tracked and gave valuesof (213 ± 139) nm/s [52] and �410–450 nm/s [53], respectively.

In this work, we probe cellular uptake and dynamics of AuNPsoptically. The particle uptake is studied with non-functionalised(citrate-capped) AuNPs revealing different trajectories and speedswhich correlate with different transport and diffusion mechanismsinside cells. Such internalised AuNPs serve as intracellular SERSprobes. We use these nanoparticle-probes taken up through theendocytotic uptake pathway to report SERS and utilise thisinformation to evaluate their ability to segregate different cellphenotypes. Further we also employed AuNPs functionalised with

Please cite this article in press as: A. Huefner et al., Methods (2014), http://dx

nuclear localisation signal peptide (NLS) as SERS probes. Thefunctionalised nanoparticle-probes give much better cellularphenotype distinction compared to non-functionalised AuNPs. Thiswork reaffirms the nanoparticle-probe based SERS methodologyfor intracellular investigations and for achieving cellulardistinction.

2. Methodology

2.1. Cellular uptake and dynamics of AuNPs inside cells

All experiments were carried out on undifferentiated (UDC) anddifferentiated (DC) SH-SY5Y cells, a human neuroblastoma cellline, cultured and maintained as described elsewhere [26]. Bothcell phenotypes are adherent and display a flat and neuronal mor-phology. Fluorescence staining with Hoechst 33343 (Invitrogen,UK) and anti-dopamine antibody (Anti-DA) (mouse, 1/1000, Milli-pore, UK) were carried out according to staining procedures de-scribed elsewhere [26,54]. Cells were grown in collagen-coatedglass bottom dishes (MatTek, US) and incubated with citrate-capped spherical AuNPs (BBinternational, UK) of different diame-ters (40, 60 and 100 nm) at a concentration of 200,000 particlesper cell independent of particle size illustrated in Fig. 1A. The up-take of particles is shown for an incubation time of 24 and 48 hin Fig. 1B–D for 40, 60 and 100 nm particles. Based on theseimages, internalisation of particles is visible. We could observeaggregates of nanoparticles inside cells, which appear as darkspots. No aggregation of AuNPs is observed in the cell culture med-ium (see Supporting Information) and therefore we believe thataggregation is induced inside the endocytotic compartments.There is no indication that non-functionalised, intracellular AuNPsare localised outside the endocytotic pathway as also observedusing transmission electron microscopy by Tkachenko et al. [37].In the images shown in Fig. 1B–D, some nanoparticles can be seenoutside cells. These are immobile aggregates which stick to thecoating of the cell culture dishes.

In order to track and characterise the motion of particles orintracellular vesicles, they have to be imaged in a time-dependentway that allows extraction of important biophysical parameterslike speed of the particle and the diffusion constant. This is com-monly performed by measuring the mean-square displacement(MSD) of a particle within a given lag time Dt. It is calculated intwo dimensions as follows:

MSDðDtÞ ¼< ðDrðDtÞÞ2 >¼< ðxðtÞ � xðt þ DtÞÞ2 þ ðyðtÞ � yðt þ DtÞÞ2 > : ð1Þ

The resulting trajectories can be classified into models describ-ing different types of motions such as anomalous subdiffusion orconfined random walk

MSDðDtÞ ¼ 4DDta ð2Þ

and directed motion as superposition of diffusion and transport

MSDðDtÞ ¼ 4DDt þ ðvDtÞ2: ð3Þ

While based on the Einstein-Stokes relation, the diffusion constantD for spherical particles subject to Brownian motion in two dimen-sions can be described as

D ¼ kBT=ð4pgrÞ ð4Þ

where kB is the Boltzmann’s constant, T is the absolute temperature,r is the particle radius and g is the fluid shear viscosity.

For tracking nanoparticles, bright field images of live cells wereobtained in trans-illumination using a 40� condenser (NA = 0.6)and a 100� oil immersion (NA = 1.4) objective with a white lightsource on a confocal microscope system (Leica, Germany). Images

.doi.org/10.1016/j.ymeth.2014.02.006

Fig. 1. Schematic of methodology (A) and images showing cellular uptake of 40 nm (B), 60 nm (C) and 100 nm (D) AuNPs after 48 h of incubation in differentiated SH-SY5Ycells. (E) The uptake of 40 nm (blue bars) and 60 nm (red bars) AuNPs into differentiated SH-SY5Y cells (n = 20 for each group) was determined by processing optical images.

A. Huefner et al. / Methods xxx (2014) xxx–xxx 3

were captured using the Leica microscope software, Leica LAS AFLite. We recorded time lapse videos of DCs incubated with100 nm AuNPs for 72 h with 120–180 frames and frame rate of1 frame/s.

Image analysis was performed using ImageJ software (Rasband,W.S., ImageJ, US National Institutes of Health, Bethesda, Maryland,USA, http://imagej.nih.gov/ij/, 1997–2012) [55]. The colour thresh-old of single images was modified manually to allow accurate rec-ognition of particles inside cells for later analysis. In order toestimate cellular particle uptake with relevance to later SERS mea-surements, the number of nanoparticles was determined by calcu-lating the surface area of cells occupied by nanoparticles dividedby the surface area A of a single nanoparticle (A = pr2). 20 cellsper sample group were analysed.

In order to characterise the motility of intracellular AuNPs,nanoparticle tracking was performed by using speckletrackerj, anadditional plugin for ImageJ software (developed by Athena’sgroup, Lehigh University, US) [56]. AuNPs were tracked for parti-cles which appeared as diffraction limited spots or larger.

2.2. SERS studies with intracellular AuNPs

2.2.1. Sample preparationFor AuNP-based SERS on UDCs and DCs, cells were grown on

13 mm, poly-L-lysine (Sigma–Aldrich, UK) coated glass cover slips(Agar Scientific, UK). As indicated in Fig. 1E, cellular particle uptakevaries with particle size. In order to make the two different particlesizes comparable, volume equivalence for the particles was chosen.Hence, 40 nm AuNPs were incubated at 675,000 particles per cellwhile 60 nm AuNPs were incubated at a concentration of 200,000particles per cell in a multi-well plate (see Fig. 1A). Each wellwas seeded with approximately 50,000 cells. Previously, we haveshown using UV–visible spectra (also see Supporting Information)that under these experimental conditions, AuNPs remain dispersedas a colloid in solution [26] which might be due to the formation ofa protein corona around the NPs in cell culture medium [57]. Asmall fraction of AuNPs adhere to the poly-L-lysine coating of thecell culture dish forming small, immobile aggregates. In order toincrease the cellular nanoparticle uptake, an incubation time ofup to 72 h was used. The cellular uptake efficiency did not pose alimiting factor for SERS measurements in our experiments. Wehave also shown earlier that the cells remain viable under theseincubation conditions and the methodology in this paper is primar-ily based on our earlier paper [26]. Thereafter, cells were washedtwice in cell culture medium in order to remove particles not takenup. After washing twice with phosphate buffered saline, cells were

Please cite this article in press as: A. Huefner et al., Methods (2014), http://dx

incubated in 4% formaldehyde for 10 min for fixation and after-wards kept in phosphate buffered saline for SERS measurements.

AuNPs functionalised with the nuclear localisation signal pep-tide (CGTG-PKKKRKV-GGK-(Flu)peptide sequence, PeptideSynthet-ics, Fareham, UK) were incubated with UDCs as well as DCs. 375 llof 40 nm AuNPs reagent (9 � 1010 particles/ml) or 385 ll of 60 nmAuNPs reagent (2.6 � 1010 particles/ml, both BBinternational, UK)were conjugated with NLS in a concentration NLS:AuNP of 100:1as described earlier [26].

2.2.2. Acquisition, processing and analysis of SERS dataA Renishaw� inVia Raman microscope with a 633 nm laser in

streamline mode and a Leica 100� (NA = 0.85) objective in combi-nation with Wire3.3 software was used to acquire spectral SERSdata sets of whole cells (one cell per data set). Generated mapshave a pixel size of 600 � 600 nm. The collection time per line ofspectra was 20 s over a spectral range from 400 to 2200 cm�1.The excitation intensity was approximately 2 � 104 W/cm2.

Spectral data was pre-processed using MATLAB R2010bemploying custom-made procedures for removal of the SERS back-ground and data set reduction as described earlier [26]. Followingdata pre-processing, data analysis was carried out as a combinationof principal component analysis (PCA) and linear discriminantanalysis (LDA) using MATLAB R2010b with IRootLab (https://code.-google.com/p/irootlab/), a graphical user interface toolbox forvibrational biospectroscopy data analysis [58,59]. As Martin et al.have described, PCA is a common technique to classify biologicalsample groups [59]. PCA, an unsupervised, multivariate data reduc-tion technique, has previously been used in context with SERSimaging [20], e.g. to achieve high diagnostic sensitivity in cancerdetection [16]. Generated SERS spectral maps require a powerfulanalysis method to reduce its dimensionality and recognise acommon pattern such as by using PCA. It generates principle com-ponent (PC) loadings and PC scores from the initial or pre-processed data. The original data and PC scores are correlated bythe PC loading, which is similar to a correlation coefficient. PC load-ings identify features (i.e. peak in a SERS spectrum) with highestimportance within the data. Therefore, PC1 loadings plots allowfor characterisation of the spectral variation within the data set.As PCA describes similar patterns which correlate to the data, theseare not necessarily those which allow for ultimate data set distinc-tion. Therefore, a method is required which finds differences with-in data sets. LDA is a supervised method for group classification. Itseeks for n-1 projections of n data sets that allow for completeseparation of those groups. LDA by itself does not facilitate datareduction. [60] Therefore, a two-stage feature extraction method

.doi.org/10.1016/j.ymeth.2014.02.006

4 A. Huefner et al. / Methods xxx (2014) xxx–xxx

(i.e. PCA-LDA) is required to allow for data reduction andclassification.

In our study, individual, pre-processed data sets (150 singlespectra per data set) were mean-centred and PCA was applied.Spectral, normalised PC1 loadings plots were used to characteriseand compare sample groups. In order to allow for group classifica-tion, PCA–LDA was applied on mean-centred, vector-normalisedindividual data sets and LD1 scores vs. LD2 scores were plottedwith data point colouration according to cell groups. Further char-acterisation of LD scores distributions was done using 1D intensitycurves. They were generated employing Kernel density estimation(bandwidth = 1000) to smooth LD scores histograms (with binsize = 50) and normalised.

3. Results and discussion

3.1. Size-dependent cellular uptake

Results from cellular uptake of nanoparticles of different sizesare shown in Fig. 1E for differentiated cells (DCs). We observedan increment in the number of AuNPs taken up with increasedincubation time from 24 to 48 h independent of particle size.Fig. 1E indicates an increase of 90% in cellular uptake of 40 nmand 51% increase for 60 nm AuNPs, during the subsequent 24 hof incubation. An increase in size of the nanoparticles leads to anincrement in the number of AuNPs taken up which is in accordancewith other studies [30,61–63]. These results reveal that cellular up-take is dependent on size and incubation time.

3.2. Motility of intracellular AuNPs

Following intracellular uptake through endocytosis, all intracel-lular nanoparticles are localised in endocytotic vesicles such asendosomes and lysosomes. These vesicles are known to be trans-ported through the cell via molecular motors [64]. Usually, brightfield images do not reveal the localisation and motility of endocy-totic vesicles within a cell due to almost negligible optical contrastcompared to the surrounding cellular structures. As AuNPs used forour studies are taken up through the endocytotic pathway, theyserve as a ‘stain’ for visualising endocytotic vesicles. Particulatesand aggregates of different sizes are seen inside cells as shown inFig. 2A. Tracking their movement (as described in Section 2.1) re-veals different types of motions which give information about cel-lular characteristics such as the viscosity/diffusion constant insideendocytotic vesicles and transport velocity. For understanding thiswe employed particle tracking analysis to see the movement ofAuNPs aggregates inside and outside SH-SY5Y cells. Using particletracking software as described in Section 2.1, 150 particle aggre-gates were randomly selected from time-lapse videos of DCs. Ascells were incubated with 100 nm AuNPs for 72 h, endocytoticvesicles appear to have a round shape and are packed with AuNPs(Fig. 2A). Particle aggregates used for tracking followed a sizedistribution as shown in Fig. 2B with an average radius of(464 ± 64) nm. The rather larger size of vesicles is advantageousfor tracking its motion as vesicles that were slightly out of focuswere still detectable. Detected tracks refer to the centre point ofthe particle aggregates.

The MSD was calculated according to Eq. (1)using MATLABR2010b. Fig. 2 shows typical traces and MSD for the four differentmotions that were observed. Fig. 2C–F show the displacements intwo dimensions (also see Supporting Information for results plot-ted with the same x and y-axis scale) and Fig. 2G–J show the appro-priate MSD vs. Dt plots. It is pointed out that our 2-D analysiswould underestimate the velocities compared to a full (3D) vecto-rial analysis. However, due to the relatively flat cells (neuronal

Please cite this article in press as: A. Huefner et al., Methods (2014), http://dx

cells) used in the study; the z-movement is likely to be minimal.In any case the general description and analytical principles wouldstill be valid with our 2-D analysis. The MSD vs. Dt plots allow foreasy categorisation of different motions. Fig. 2C and G show anexample for an isotropic random walk. Fitting Eq. (2)to the MSDvs. Dt plot in Fig. 2G reveals a diffusion constant, Ddiffusion =(0.70 ± 0.15) lm2/s of the particle aggregate inside an endocytoticvesicle within the cytoplasm as well as the time exponent,a = 1.03 ± 0.06. a = 1 in Eq. (2)implies that this motion correspondsto Brownian motion of particulates inside cells. An example of apartially confined random walk is shown in Fig. 2D and H. This mo-tion is characterised by alternating regions of almost linearlyincreasing MSD and plateaus indicating that the movementchanges from random walk to confined motion. An oscillatingMSD over Dt as visible in Fig. 2I is characteristic for an oscillatingmotion as confirmed by the track in Fig. 2E. This motion suggeststhe adsorption of particles to the membrane inside vesicles as alsodescribed by Jin et al. [65]. Fig. 2F represents a confined motion/transport of a vesicle within the cell superposing the diffusion ofthe vesicle within the cytoplasm. Typical trajectories follow non-uniform linear motions; their MSD over Dt (Fig. 2J) increases qua-dratically as described by Eq. (3). Fitting Eq. (3)(red line in Fig. 2H)reveals an exemplary transport velocity of vtransport = (39 ± 4) nmand a cytoplasmic diffusion constant during active transport ofDtransport = (6.6 ± 0.6) � 10�3 lm2/s. The contribution of activetransport is much higher than that of free diffusion which suggestsa binding of the vesicle to the cytoskeletal filament network viamolecular motors. Values for the diffusion constant and the vesiclevelocity are in agreement with results from Levi et al. [50] and Cas-pi et al. [51].

Generally, the track of one particle shows intervals of differenttypes of motions impeding a clear distinction between them. Just afew tracks exclusively show only linear motion or active transport.To translate findings from Fig. 2J to an average transport velocity,motion tracks (n = 23) showing active transport have been chosenas summarised in the MSD-Dt log–log plot in Fig. 3A. The runningaverage of MSD was calculated according to Eq. (1), followed bycalculating the diffusion constant D = MSD/(4Dt), where Dt waschosen to be smaller than a quarter of the total number of datapoints [50]. This common methodology is based on the unbiasedassumption that the dominant motion (active transport over diffu-sion) contributes the most to the overall motion of the particle. Thehistogram of –logD is presented in Fig. 3B showing a double peakGaussian distribution. A multi peak fit (red line) reveals the peakpositions as 0.20 ± 0.03 and 1.92 ± 0.05 equivalent to diffusion con-stants of D1 = 0.6 � 10�1 lm2/s and D2 = 10�2 lm2/s, respectively.The smaller distribution around D1 corresponds to the diffusionconstant of AuNP aggregates themselves inside endocytotic vesi-cles in the cytoplasm. The second, much higher distribution aroundD2 corresponds to the diffusion constant of AuNP aggregates insideendocytotic vesicles while actively transported by molecularmotors attached to the filament network of the cytoskeleton. Thisexplains why D1 is larger than D2. This double distribution is indic-ative of the diffusion constant for active transport as well as for dif-fusion inside the intracellular environment and has been describedbefore in particle tracking experiments with quantum dot-conju-gated prion proteins inside yeast cells by Tsuji et al. [66]. Further-more, Fig. 3C summarises the velocity distribution of activetransport, showing a dominant peak at (36 ± 2) nm/s followed bytwo minor peaks at (182 ± 24) nm/s and at (251 ± 39) nm/s deter-mined by local Gaussian fits (data not shown). This indicates thatactive transport with a velocity of �30–40 nm/s is the dominantprocess, but faster motions could also be observed with valuescomparable to velocities of endocytic traffic observed by Toshimaet al. [52]. Aggregates of AuNPs as tracked in our study show ahigher density and average sizes that are in excess compared to

.doi.org/10.1016/j.ymeth.2014.02.006

-2 -1 0 1-3

-2

-1

0

1

2

3

x-displacement (μm)

y-di

spla

cem

ent (

μm)

0 0.2 0.4 0.6-0.4

-0.2

0

0.2

0.4

x-displacement (μm)y-

disp

lace

men

t (μm

)

0 20 40 60 800.06

0.08

0.1

0.12

0.14

0.16

Δt (s)

<(Δr

( Δt))

2 > ( μ

m2 )

0 1 2 3-4

-3

-2

-1

0

1

x-displacement (μm)

y-di

spla

cem

ent (

μm)

0 20 40 600

2

4

6

8

Δt (s)<(

Δr( Δ

t))2 >

( μm

2 )

0 20 40 600

2

4

6

8

Δt (s)

<(Δr

( Δt))

2 > ( μ

m2 )

-8 -6 -4 -2 00

5

10

15

x-displacement (μm)

y-di

spla

cem

ent (

μm)

0 10 200

20

40

60

80

Δt (s)

<(Δr

( Δt))

2 > ( μ

m2 )

C D E F

G H I J

250 300 350 400 450 500 550 600 650 7000

5

10

15

20

25

Radius (nm)

Cou

nts

BA

10μm

Fig. 2. (A) The bright field image shows intracellular nanoparticle aggregates after 72 h of incubation. Those aggregates used for motility tracking showed an average radius of(464 ± 64) nm (B). Tracking analysis revealed different types of motions, where (C–F) show the trajectories and (G–J) the corresponding MSD vs. Dt plots. The followingmotions are examples of motions commonly observed: random walk of particle aggregate inside vesicle (C, G), partially confined random walk (D, H), oscillation of particlesdue to confinement to membranes (E, I), transport process/confined motion (F, J). Red lines (G, J) are fittings to the MSD plots revealing the diffusion constant inside vesicles tobe Ddiffusion = (0.70 ± 0.15) lm2/s (G) and the typical transport velocity vtransport = (39 ± 4) nm/s (J). Plots in C–F have different scales to highlight the particle trajectories. Forplots with a uniform x-scale see Fig. S2 in the supporting information.

A. Huefner et al. / Methods xxx (2014) xxx–xxx 5

those used in other studies [46–49]. This can result in increasedsteric hindrance during active transport in addition to an increasedload ratio on the molecular motors that ultimately causes theirmotion to be slowed down as been shown by Coppin et al. [67].They observed the slow-down and final stall of kinesin motors un-der opposing loads of 5 pN together with dissociating of kinesinmotors from microtubules.

In order to actively move aggregates of AuNPs through a med-ium/cell, the pulling force has to exceed the drag force F = 6pgrv(r is the radius of the spherical particle and v its velocity, g is thedynamic viscosity) under the idealised assumption of hydrody-namic regime for the friction and under idealised geometries. Eq.(4)connects the diffusion D with the viscosity g resulting in the re-quired drag force F = vkBT/D with kB is the Boltzmann’s constantand T the temperature. Using our experimental results vtransport

and Dtransport of active transport as calculated in Fig. 2J reveals a re-quired drag force of 0.1 pN. As estimated under idealised condi-tions, the required drag force for actively transported AuNPaggregates with a size of �1 lm might under real conditions even-tually reach the range of described stall forces connected with aslow-down of transport motions as observed by Coppin et al. [67].

Following the tracking of AuNPs inside and outside cells, themotility of actively transported particles was characterised further

Please cite this article in press as: A. Huefner et al., Methods (2014), http://dx

(Fig. 3A–C). Trajectories of all 150 tracked AuNP aggregates aresummarised in a –logD plot in Fig. 3D. Multiple Gaussian peakswere fitted to the distribution which reveal a shoulder at�0.46 ± 0.10 as well as peaks at 0.48 ± 0.03 and 1.70 ± 0.07 corre-sponding to D values of �10�1, �10�1 and �10�2 lm2/s, respec-tively. These diffusion constants represent different conditions.The first D value represents AuNP aggregates diffusing outsidethe cell indicated by the high diffusion constant. This has been de-scribed in particle tracking experiments in biological buffers inother studies before [66]. The latter values are related to activetransport corresponding to diffusion inside endocytotic vesiclesand active transport whilst being attached to molecular motors.Furthermore, we observed outliers for 3 < �logD < 4 correspondingto D = 10�4–10�3 lm2/s. Such a low diffusion constants are charac-teristic to oscillating motions (Fig. 2E and I) corresponding to par-ticles being confined to membranes.

In summary, particle tracking used for studying the motility ofparticle aggregates helped in understanding the location (inside oroutside cell using diffusion coefficient) and behaviour (e.g. aggre-gation inside intracellular, endocytotic vesicles) of intracellularAuNPs. Knowing these characteristics in combination with quanti-tative estimations of particle uptake (as investigated in Section 3.1)demonstrate that AuNPs are suitable intracellular probes for

.doi.org/10.1016/j.ymeth.2014.02.006

0 20 40

10-2

100

102

Δt (s)

<(Δr

( Δt))

2 > ( μ

m2 )

0 5000

100

200

300

400

Velocity (nm/s)

Cou

nts

0 2 40

100

200

300

-log(D)C

ount

s

0 2 40

500

1000

1500

-log(D)

Cou

nts

C

A B

D

Fig. 3. Particle aggregates following linear tracks (n = 23) were chosen and MSD vs.Dt (A), �logD histogram (B) as well as the velocity distribution (C) are plotted. 150vesicles were tracked over time intervals between 40 and 180 s. MSD wascalculated for Dt = 10 s for each track shown in a �logD histogram (C) withmultiple Gaussian peaks (red line) fitted to the distribution with a shoulder at(�0.46 ± 0.10), peaking at (0.48 ± 0.03) and (1.70 ± 0.07) as well as outlier for3 < �logD < 4. They represent the diffusivity of particles free in the cell culturemedium (D = 100–10�1 lm2/s), diffusivity inside cellular organelles of the endocy-totic pathway (D = 10�1 lm2/s), diffusivity during active transport (D = 10�2 lm2/s),as well as oscillating motion (D = 10�4–10�3 lm2/s).

6 A. Huefner et al. / Methods xxx (2014) xxx–xxx

techniques taking advantage of particle uptake. SERS exploits theseproperties ideally as it uses intracellular aggregates of AuNPs asnanoantennas to report their local chemical environment. Our re-sults exploiting the natural uptake of AuNPs to act as intracellularSERS probes are detailed below.

3.3. Cellular imaging using SERS

While incubation of cells with AuNPs, particles are gradually ta-ken up through the endocytotic pathway inside the cell. Intracellu-lar AuNPs are likely to form aggregates inside of endosomes/lysosomes. Exemplarily this is shown for an undifferentiated cell(UDC) incubated with 40 nm AuNPs for 72 h in Fig. 4A. While AuNPaggregates of different sizes are distributed all over the cell thenuclear region, highlighted by a white ellipse, is clearly omitted.Intense SERS signals could only be achieved from those areasshowing aggregates.

Raman scanning measurements gave a spectrum for every pixelinterrogated within the field of interest, usually, in our case, a sin-gle cell. The intensity of single peaks or peak regions, assigned to aparticular molecular vibration(s), within the field of interest can beused to generate pseudo-colour maps representing their distribu-tion within the sample. This allows for tracking several differentmolecules, i.e. proteins or lipids, located in the close vicinity ofour ‘nanoparticle-probes’ after a single scan of the sample withoutthe need for molecule-specific stains and/or markers. This makesSERS hugely advantageous over other molecular imaging tech-niques. Fig. 4B–C shows the molecular distribution of proteins(Fig. 4B) and lipids (Fig. 4C) of the sample cell shown in Fig. 4A.Characteristic peptide and protein bands in SERS are amide I(1600–1700 cm�1) and amide III (1200–1400 cm�1) bands. Fur-thermore, lipids show characteristic SERS peaks at 1116 (C–Cstretch), 1260 (C–H stretch), 1300 (C–H2 twist) and 1440 cm�1

(C–H2 bend). [68] Exemplary SERS spectra from the maps gener-ated in Fig. 4B–C are presented in Fig. 4D showing some of the

Please cite this article in press as: A. Huefner et al., Methods (2014), http://dx

mentioned protein and lipid peaks. Its high molecular resolutionmakes SERS a suitable method for sensing biomolecular changes.In this context, we demonstrate that intracellular nanoparticles-based SERS allows the distinction of closely related cell (pheno)-types which differ in their biochemical makeup.

3.4. Intracellular SERS imaging with non-functionalised AuNPs

The use of optical and fluorescence microscopy has led to manydevelopments in identification of cellular structures. However, dueto the lack of chemical distinction ability, non-invasive and label-free classification of closely related cell phenotypes remains oneof the big challenges. We first employed bare 40 nm AuNPs volun-tarily internalised by cells through the endocytotic pathway asintracellular SERS nanosensors in UDCs (n = 12) and DCs (n = 26);closely related neuronal cells showing similar morphologies [69].These citrate capped (non-functionalised) nanoparticle-probes(also see Sections 3.1 and 3.2) are taken up entirely through theendocytotic pathway. Hence, these SERS nanoparticle-probesshould sense the ingredients of this cellular ‘digestive system’ ofthe cell.

Single cell SERS maps were acquired from cells of each cellgroup (UDCs and DCs), data sets pre-processed (see Section 2.2.2)and PCA was performed. In Fig. 4E resulting PC1 loadings for allUDCs (red lines, middle) and DCs (blue lines, top) as well as theiraverage loadings (bottom) are shown. Comparing PC1 loadings ofboth cell groups, small differences between the cell group averageloadings can be seen in the range of 1100–1750 cm�1. In order toevaluate if these differences between the average loadings arecaused by general divergence within samples of the same groupor between the cell groups, corresponding standard deviations(STD) within all PC1 loadings of each cell group (top spectrum inFig. 4F) as well as the difference spectrum of the average PC1 load-ings of UDCs and DCs (bottom) are given in Fig. 4F. The STD of sin-gle PC1 loadings between 1100 and 1750 cm�1 partly show valuesabove 5%. In particular, between 1100–1200 cm�1 (STD of 8%) andwithin the amide I band (i.e. 1300–1380 cm�1, 1470–1510 cm�1

and 1675–1710 cm�1) the STD increases up to 12%. Therefore,the difference spectrum in Fig. 4F (bottom) rather appears to besubject to the variance of data sets within one cell group thanthe variance within the both cell groups. In this case PCA by itselfdid not allow for cell group segregation, hence, PCA–LDA was ap-plied and the resulting LD1 vs. LD2 scores 2D scatter plot is shownin Fig. 4G. Corresponding 1D intensity plots (see Section 2.2.2) ofthe LD1 and LD2 score distributions are shown alongside for UDCs(red) and DCs (blue). As expected from PCA results, the scatter plotshows a significant overlap between both cell groups. Only the 1Dintensity curves reveal more detailed characteristics of the datapoint distribution. For UDCs, the LD1 scores intensity distributionis Gaussian and peaks at �0.12. The LD1 scores distribution ofDCs peaks at 0.09 and has a shoulder at around �0.04. LD2 scoresdistributions are Gaussian peaking at 0.22 and �0.4 for UDCs andDCs, respectively. These results in combination with the PC1 load-ings confirm great similarities between the cell groups. In particu-lar, the distribution of LD2 scores of both cell groups are very closetogether. Also, LD1 scores of UDCs overlap broadly with those ofDCs for values <0. Nevertheless, PCA–LDA revealed that slight dif-ferences are also present. The LD1 scores double-peak distributionof DCs has one peak/shoulder completely overlapping with UDCs,but also shows a peak separated from the other cell group, indicat-ing that some molecular features are only present in DCs. Cellulardifferentiation in SH-SY5Y cells was induced using staurosporine,an inhibitor of various protein kineases [70,71]. Staurosporine ini-tially causes a change of the cellular metabolism as observed byDeshmukh and Johnson [72] even though exact mechanisms in-volved are not known yet [70,71]. These metabolic changes on

.doi.org/10.1016/j.ymeth.2014.02.006

600 1000 1400 1800

0

0.5

1

1.5

2

2.5

Raman Shift (cm-1)

PC

1 lo

adin

gs

average PC1 loadings

DCsUDCs

0 0.5-1

-0.5

0

0.5

1

LD2

scor

es

-1 0 10

0.5

1

LD1 scores

Nor

mal

ised

in

tens

ity

600 1000 1400 18000

2000

4000

6000

Raman Shift (cm-1)

Cou

nts

(a.u

.)

600 1000 1400 18000

0.05

0.1

Raman Shift (cm-1)S

tand

. Dev

.

600 1000 1400 1800

-0.05

0

0.05

0.1

Raman Shift (cm-1)

Diff

eren

ceS

pect

rum

A B C D

GFE

Fig. 4. (A) Bright field image of a UDC after 72 h of incubation with 40 nm AuNPs showing many intracellular particle aggregates outside the cell nucleus as highlighted with awhite ellipse (scale bar: 10 lm). Corresponding SERS maps (B–C) revealing the intracellular distribution of proteins ((B) green: C–C/C–N stretch, yellow: C–H deformation,cyan: C–H3 deformation) and lipids ((C) magenta: C–H2 twist, red: C@C stretch). The SERS images were generated by scanning with a pixel size of 600 � 600 nm. (D) Five,background subtracted spectra (D) from different intracellular positions exemplary illustrate the molecular variance recorded by our ‘nanoantennas’ inside cells. (E–F)Principal component loadings of single cell SERS data sets from UDCs (red) and DCs (blue) incubated with non-functionalised 40 nm AuNPs for up to 72 h. Single cell PC1loadings (E) as well as cell group average (bottom, (E)), standard deviation (top, (F)) and the difference spectrum of the cell group averages of UDCs and DCs (bottom, (F)) showthat differences in the average PC1 loadings of cell groups are rather caused by divergences within the cell group. (G) PCA–LDA revealing LD1 vs. LD2 scatter plot withcorresponding 1D intensity plots alongside. Even though LD scores show overlap between the cell groups, the LD1 scores intensity curve of UDCs (red) and DCs (blue) showsclear distinction due to its shape.

A. Huefner et al. / Methods xxx (2014) xxx–xxx 7

the one hand drive the differentiation process [73,74] and on theother hand cause shifts between cellular glycolysis and oxidativephosphorylation due to modified cellular metabolite level and re-dox state [75]. These intracellular developments might cause thecellular uptake needs of the cell to change slightly which we wereable to detect using AuNP-based SERS of endocytotic vesicles.However, the induced metabolic changes are subtle. Thus althoughwith the non-functionalised probes SERS active AuNPs inside the‘digestive’ endocytotic pathway some differences are identifiedcorresponding to metabolic changes induced by differentiationbut distinction of the two cell types was only achieved using thepowerful, two-step analysis method of PCA–LDA. PCA alone didnot allow for cell group distinction.

3.5. Functionalised SERS nanoparticle-probes

Functionalised nanoparticle-probes have been used to targetintracellular molecules [76], structures [77] and organelles suchas the mitochondria [78], cytoplasm [34] and the cell nucleus[34,44,79]. One of the challenges with this approach is to enablethe escape of nanoparticles from endocytotic vesicles [30,77].Studies have shown that nanoparticle functionalisation achievescytoplasmic localisation of the cargo using target proteins and pep-tides such as the cell penetrating peptides (CPPs) [80], SV40 large Tantigen [81], HIV-1 Tat peptide [34] as well as adenoviral NLS [44].Some of these proteins also show nuclear localisation although themajority of internalised particles are still found in endosomes andthe cytoplasm [44]. We have used the SV-40 large T nuclear local-isation signal bound to 40 nm AuNPs for their successful transloca-tion into the cell nucleus although most of them localise into thecytoplasm as well as endosomes/lysosomes [26]. NLS functional-ised AuNPs (NLS-AuNPs) localise into the cytoplasm by escapingthe endocytotic pathway and can also translocate into the nucleusas shown in our previous work [26]. Exemplar images of an

Please cite this article in press as: A. Huefner et al., Methods (2014), http://dx

undifferentiated cell incubated with NLS-AuNPs for 72 h are shownin Fig. 5A–C.

In the current work we mainly focussed on AuNPs outside thenucleus. The SERS spectra of UDCs (n = 16) and DCs (n = 11) incu-bated with NLS-AuNPs for 72 h were characterised using PCA aswell as PCA–LDA with regards to their cellular segregation abilityas shown in Fig. 5. Single cell PC1 loadings of UDCs (red, top) andDCs (blue, middle) as well as cell group averages (bold blue andred line, bottom) are presented in the top plot in Fig. 5D. PC1 load-ings of DCs show a very small divergence. UDCs show a greaterdivergence, in particular between 1200 and 1600 cm�1 indicatinga variation in the amide III band (1200–1400 cm�1) assigned toproteins. Nevertheless, average PC1 loadings show very distinctshapes allowing for clear cell group segregation. The differencespectrum (black line) in the bottom plot of Fig. 5D reveals shiftsin the protein rich regions around 1100–1200 cm�1 (peaks at1160 cm�1 and 1178 cm�1 for DCs and UDCs, respectively) and1500–1700 cm�1 (peaks at 1575 cm�1 and 1585/1605 cm�1 forDCs and UDCs, respectively). Also, there is a change in intensity ra-tios between various peaks in the PC1 average loadings. Mentionedpeaks are assigned to the C–N stretching and the C–H bending inthe amino acids tyrosine and phenylalanine [11,82]. DCs are com-monly used as a model of dopaminergic neurons for Parkinson’sDisease research [69]. Dopamine beta hydroxylase, an enzyme in-volved into the synthesis of the neurotransmitters dopamine andnorepinephrine, has been shown to be present in DCs [83]. Aminoacid peaks found in DCs are also present in spectra of dopamineand norepinephrine [84] suggesting the presence of these neuro-transmitters in DCs and their contribution to SERS spectra. To con-firm the validity of this hypothesis, UDCs and DCs werefluorescently immuno-stained for dopamine and the results areshown in Fig. 6. Besides nuclear staining (Fig. 6A and D), fluores-cence images for dopamine (Fig. 6B and E) and merged images(Fig. 6C and D) show the presence of dopamine in some DCs

.doi.org/10.1016/j.ymeth.2014.02.006

0

0.5

1

1.5

2

2.5

Raman Shift (cm-1)

PC

1 lo

adin

gs

average PC1 loadings

DCsUDCs

600 1000 1400 1800

-0.2

0

0.2

Raman Shift (cm-1)

Diff

eren

ceSp

ectru

m 0 0.5

-1

-0.5

0

0.5

1

LD2

scor

es-2 -1 0 1 20

0.5

1

LD1 scores

Nor

mal

ised

in

tens

ity

D E

A CB

Fig. 5. (A) Bright field image of a UDC after 72 h of incubation with 40 nm NLS functionalised AuNPs showing many intracellular particle aggregates outside as well as a fewinside the cell nucleus as highlighted with a white ellipse (scale bar: 10 lm). Corresponding SERS maps (B–C) revealing the intracellular distribution of proteins ((B) green: C–C/C–N stretch, yellow: C–H deformation, cyan: C–H3 deformation) and nucleic acids ((C) red: DNA bands at 670, 830, 1375 and 1580 cm�1 and green: RNA peak at 815 cm�1).The SERS images were generated by scanning with a pixel size of 600 � 600 nm. PCA (D) and PCA–LDA (E) of intracellular NLS-AuNPs in UDCs (red) and DCs (blue). The topfigure in (D) shows the PC1 loadings plot for single cells (top for DCs, middle for UDCs) and cell group average (bottom). The difference spectrum (bottom plot in (D)) revealssignificant difference between UDCs and DCs in the amino acid regions around 1100–1200 cm�1 and 1500–1700 cm�1. Peaks in these region show an amplitude of 20–30%compared to the average PC1 loadings. (E) LD1 vs. LD2 scatter plot with corresponding 1D intensity curves showing a double peak distribution of LD1 and LD2 scores for DCswith one peak overlaying the distribution of UDCs. This indicates the existence of in common intracellular content as well as molecular dissimilarities allowing for clearcellular distinction.

UD

Cs

DC

s

Hoechst Anti-DA Merge

A B C

FED

Fig. 6. Hoechst 33342 (A, D)/Anti-dopamine antibody (B, E) double staining of UDCs (top row) and DCs (bottom row) for visualisation of the intracellular dopaminedistribution. Merged fluorescence image of UDCs (C) shows only fluorescence background, in contrast to that of DCs (F) which proves the presence of intracellular dopamine.

8 A. Huefner et al. / Methods xxx (2014) xxx–xxx

(Fig. 6D–F) in contrast to UDCs (Fig. 6A–C). Overall these differ-ences reflect in the PC1 loadings (Fig. 5D) and allow for distinct cellgroup segregation.

Further, however, we performed LDA on the PCA results. Theseare shown in the LD1 vs. LD2 scatter plot and corresponding

Please cite this article in press as: A. Huefner et al., Methods (2014), http://dx

average 1D intensity curves in Fig. 5E (see Supporting Informationfor single cell 1D intensity curves of UDCs and DCs). DCs show adouble peak distribution for LD1 and LD2 scores peaking at�0.26/0.44 and �0.09/0.09, respectively. For UDCs, LD1 and LD2scores show an almost Gaussian-shaped curve with slight negative

.doi.org/10.1016/j.ymeth.2014.02.006

A. Huefner et al. / Methods xxx (2014) xxx–xxx 9

skew for LD1 and slight positive skew for LD2. They peak at �0.40and 0.05 for LD1 and LD2 scores, respectively. Functionalisation ofAuNPs results in a smaller overlap of LD1/LD2 scores of both cellgroups. Negative LD1 scores of DCs overlap with UDCs suggestingthe common molecular features of these cell groups. However,the majority of the data points of the LD1 scores of DCs reveal po-sitive values and are free of overlap with UDCs demonstrating thepresence of significant, molecular dissimilarities. Therefore, LD1scores allow for complete cell group classification.

Comparing these results to intracellular SERS of non-functional-ised AuNPs, an enormous improvement in group classification wasachieved by considering spectra from the cytoplasm in addition tothose from endosomes/lysosomes. Not only a very clear segrega-tion could be accomplished using PCA–LDA but also potentialintracellular changes induced by cell differentiation were visual-ised using PC1 loadings with functionalised AuNPs.

4. Conclusion

In this study, we have investigated the uptake of AuNPs andtheir use as suitable nanoparticle-sensors for intracellular SERSimaging. Taken up and after being processed through the endocy-totic pathways (endosomes/lysosomes), their motility within thispathway was investigated using particle tracking and revealedthe intracellular diffusion coefficients as well as characteristictransport velocities of endocytotic vesicles. Understanding NP up-take is of great relevance to intracellular SERS and allows probingthe ‘digestive’ system of the cell. Insight into metabolic and intra-cellular dissimilarities of different cell (pheno)types was gained bystudying SERS signals from the cytoplasm and the endocytotic ves-icles. Intracellular SERS allowed for segregation of closely relatedcell phenotypes using a combination of principle component andlinear discriminant analysis although functionalised AuNPs werefound to be better than non-functionalised AuNPs. In comparisonto UDCs, DCs showed a significant presence of specific amino acidsinvolved in the synthesis of neurotransmitters such as dopamineand norepheneprine suggesting dopamine beta hydroxylase activ-ity. Furthermore, PC loadings as well as PCA–LDA allowed for clearphenotype distinction. This work therefore highlights that AuNPsbased SERS imaging could be of great importance for biomedicalresearch involving molecular characterisation and intracellularimaging of different cell types.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.ymeth.2014.02.006.

References

[1] E. Boisselier, D. Astruc, Chem. Soc. Rev. 38 (6) (2009) 1759–1782.[2] M.J. Sailor, J.-H. Park, Adv. Mater. 24 (28) (2012) 3779–3802.[3] J. Suh, M. Dawson, J. Hanes, Adv. Drug Delivery Rev. 57 (1) (2005) 63–78.[4] S. Schlücker, ChemPhysChem 10 (9–10) (2009) 1344–1354.[5] Nie and Emory, Science 275 (5303) (1997) 1102–1106.[6] M.K. Hossain, A.O. Yukihiro, Curr. Sci. India 97 (2) (2009) 192–201.[7] A. Campion, P. Kambhampati, Chem. Soc. Rev. 27 (4) (1998) 241–250.[8] A.M. Schwartzberg, C.D. Grant, A. Wolcott, C.E. Talley, T.R. Huser, R. Bogomolni,

J.Z. Zhang, J. Phys. Chem. B 108 (50) (2004) 19191–19197.[9] J. Kneipp, H. Kneipp, B. Wittig, K. Kneipp, Nanomedicine 6 (2) (2010) 214–226.

[10] S. Mahajan, J. Richardson, T. Brown, P.N. Bartlett, J. Am. Chem. Soc. 130 (46)(2008) 15589–15601.

[11] J. Kneipp, H. Kneipp, M. McLaughlin, D. Brown, K. Kneipp, Nano Lett. 6 (10)(2006) 2225–2231.

[12] L. Li, T. Hutter, U. Steiner, S. Mahajan, Analyst 138 (16) (2013) 4574–4578.[13] K. Faulds, W.E. Smith, D. Graham, R.J. Lacey, Analyst 127 (2) (2002) 282–286.[14] A. Samanta, K.K. Maiti, K.-S. Soh, X. Liao, M. Vendrell, U.S. Dinish, S.-W. Yun, R.

Bhuvaneswari, H. Kim, S. Rautela, J. Chung, M. Olivo, Y.-T. Chang, Angew.Chem. Int. Ed. 50 (27) (2011) 6089–6092.

Please cite this article in press as: A. Huefner et al., Methods (2014), http://dx

[15] X. Li, T. Yang, J. Lin, J. Biomed. Opt. 17 (3) (2012) 037003–037005.[16] S. Feng, J. Lin, Z. Huang, G. Chen, W. Chen, Y. Wang, R. Chen, H. Zeng, Appl.

Phys. Lett. 102 (4) (2013) 043702–043704.[17] J.A. Kim, C. Åberg, A. Salvati, K.A. Dawson, Nat. Nanotechnol. 7 (1) (2012) 62–

68.[18] S. Abalde-Cela, P. Aldeanueva-Potel, C. Mateo-Mateo, L. Rodriguez-Lorenzo, R.

Alvarez-Puebla, L.M. Liz-Marz, J. R. Soc. Interface 7 (Suppl. 4) (2010) S435–450.[19] K. Kneipp, H. Kneipp, I. Itzkan, R.R. Dasari, M.S. Feld, J. Phys. Condens. Matter

14 (18) (2002) R597.[20] M.F. Escoriza, J.M. VanBriesen, S. Stewart, J. Maier, Appl. Spectrosc. 61 (8)

(2007) 812–823.[21] J. Kang, H. Gu, Probing of Cancer Cell Apoptosis by Sers and Lscm, in: F.

Amzajerdian, C.-Q. Gao, T.-Y. Xie (Eds.), SPIE, 2009 (first ed.).[22] W. Xie, L. Wang, Y. Zhang, L. Su, A. Shen, J. Tan, J. Hu, Bioconjugate Chem. 20 (4)

(2009) 768–773.[23] K. Kneipp, A.S. Haka, H. Kneipp, K. Badizadegan, N. Yoshizawa, C. Boone, K.E.

Shafer-Peltoer, J.T. Motz, R.R. Dasari, M.S. Feld, Appl. Spectrosc. 56 (2) (2002) 5.[24] K. Ock, W.I. Jeon, E.O. Ganbold, M. Kim, J. Park, J.H. Seo, K. Cho, S.-W. Joo, S.Y.

Lee, Anal. Chem. 84 (5) (2012) 2172–2178.[25] C. Matthäus, T. Chernenko, J.A. Newmark, C.M. Warner, M. Diem, Biophys J 93

(2) (2007) 668–673.[26] A. Huefner, W.-L. Kuan, R.A. Barker, S. Mahajan, Nano Lett. (2013).[27] P. Candeloro, L. Tirinato, N. Malara, A. Fregola, E. Casals, V. Puntes, G.

Perozziello, F. Gentile, M.L. Coluccio, G. Das, C. Liberale, F. De Angelis, E. DiFabrizio, Analyst 136 (21) (2011) 4402–4408.

[28] C.M. Feldherr, D. Akin, J. Cell Biol. 111 (1) (1990) 1–8.[29] J. Lin, R. Chen, S. Feng, Y. Li, Z. Huang, S. Xie, Y. Yu, M. Cheng, H. Zeng, Biosens.

Bioelectron. 25 (2) (2009) 388–394.[30] R. Lévy, U. Shaheen, Y. Cesbron, V. Sée, Nano Rev. 1 (2010) 4889.[31] T.M. Klein, E.D. Wolf, R. Wu, J.C. Sanford, Nature 327 (6117) (1987) 70–73.[32] W. Liang, J.K.W. Lam, Endosomal Escape Pathways for Non-Viral Nucleic Acid

Delivery Systems, InTech, 2012.[33] A. Verma, F. Stellacci, Small 6 (1) (2010) 12–21.[34] J.M. de la Fuente, C.C. Berry, Bioconjugate Chem. 16 (5) (2005) 1176–1180.[35] S. Zhang, J. Li, G. Lykotrafitis, G. Bao, S. Suresh, Adv. Mater. 21 (4) (2009) 419–

424.[36] E.C. Cho, Q. Zhang, Y. Xia, Nat. Nanotechnol. 6 (6) (2011) 385–391.[37] A.G. Tkachenko, H. Xie, Y. Liu, D. Coleman, J. Ryan, W.R. Glomm, M.K. Shipton,

S. Franzen, D.L. Feldheim, Bioconjugate Chem. 15 (3) (2004) 482–490.[38] I. Lynch, T. Cedervall, M. Lundqvist, C. Cabaleiro-Lago, S. Linse, K.A. Dawson,

Adv. Colloid Interface 134–135 (2007) 167–174.[39] A. Albanese, P.S. Tang, W.C.W. Chan, Annu. Rev. Biomed. Eng. 14 (1) (2012) 1–

16.[40] T. Mironava, M. Hadjiargyrou, M. Simon, V. Jurukovski, M.H. Rafailovich,

Nanotoxicology 4 (1) (2010) 120–137.[41] A. Panariti, G. Miserocchi, I. Rivolta, Nanotechnol. Sci. Appl. 2012 (5) (2012)

87–100.[42] H.T. McMahon, E. Boucrot, Nat. Rev. Mol. Cell Biol. 12 (8) (2011) 517–533.[43] A. Albanese, W.C.W. Chan, ACS Nano 5 (7) (2011) 5478–5489.[44] A.G. Tkachenko, H. Xie, D. Coleman, W. Glomm, J. Ryan, M.F. Anderson, S.

Franzen, D.L. Feldheim, J. Am. Chem. Soc. 125 (16) (2003) 4700–4701.[45] R.D. Vale, Cell 112 (4) (2003) 467–480.[46] D. Cai, K.J. Verhey, E. Meyhöfer, Biophys. J. 92 (12) (2007) 4137–4144.[47] D.S. Friedman, R.D. Vale, Nat. Cell Biol. 1 (5) (1999) 293–297.[48] S. Courty, C. Luccardini, Y. Bellaiche, G. Cappello, M. Dahan, Nano Lett. 6 (7)

(2006) 1491–1495.[49] V. Levi, E. Gratton, Cell Biochem. Biophys. 48 (1) (2007) 1–15.[50] V. Levi, Q. Ruan, E. Gratton, Biophys. J. 88 (4) (2005) 2919–2928.[51] A. Caspi, O. Yeger, I. Grosheva, A.D. Bershadsky, M. Elbaum, Biophys. J. 81 (4)

(2001) 1990–2000.[52] J.Y. Toshima, J. Toshima, M. Kaksonen, A.C. Martin, D.S. King, D.G. Drubin, Proc.

Natl. Acad. Sci. USA 103 (15) (2006) 5793–5798.[53] Š. Bálint, I. Verdeny Vilanova, Á. Sandoval Álvarez, M. Lakadamyali, Proc. Natl.

Acad. Sci. USA 110 (9) (2013) 3375–3380.[54] W.-L. Kuan, E. Poole, M. Fletcher, S. Karniely, P. Tyers, M. Wills, R.A. Barker, J.H.

Sinclair, J. Exp. Med. 209 (1) (2012) 1–10.[55] C.A. Schneider, W.S. Rasband, K.W. Eliceiri, Nat. Methods 9 (7) (2012) 671–675.[56] Matthew.B. Smith, Biophys. J. 101 (7) (2011) 1794–1804.[57] M. Chanana, P. Rivera Gil, M.A. Correa-Duarte, L.M. Liz-Marzán, W.J. Parak,

Angew. Chem. Int. Ed. 52 (15) (2013) 4179–4183.[58] J. Trevisan, P.P. Angelov, A.D. Scott, P.L. Carmichael, F.L. Martin, Bioinformatics

29 (8) (2013) 1095–1097.[59] F.L. Martin, J.G. Kelly, V. Llabjani, P.L. Martin-Hirsch, I.I. Patel, J. Trevisan, N.J.

Fullwood, M.J. Walsh, Nat. Protoc. 5 (11) (2010) 1748–1760.[60] J. Yang, J.-Y. Yang, Lect. Notes Comput. Sci. 36 (2) (2003) 563–566.[61] B.D. Chithrani, A.A. Ghazani, W.C.W. Chan, Nano Lett. 6 (4) (2006) 662–668.[62] F. Lu, S.-H. Wu, Y. Hung, C.-Y. Mou, Small 5 (12) (2009) 1408–1413.[63] D.B. Chithrani, Mol. Membr. Biol. 27 (7) (2010) 299–311.[64] J.L. Ross, M.Y. Ali, D.M. Warshaw, Curr. Opin. Cell Biol. 20 (1) (2008) 41–47.[65] H. Jin, D.A. Heller, M.S. Strano, Nano Lett. 8 (6) (2008) 1577–1585.[66] T. Tsuji, S. Kawai-Noma, C.-G. Pack, H. Terajima, J. Yajima, T. Nishizaka, M.

Kinjo, H. Taguchi, Biochem. Biophys. Res. Commun. 405 (4) (2011) 638–643.[67] C.M. Coppin, D.W. Pierce, L. Hsu, R.D. Vale, Proc. Natl. Acad. Sci. USA 94 (16)

(1997) 8539–8544.[68] H. Wu, J.V. Volponi, A.E. Oliver, A.N. Parikh, B.A. Simmons, S. Singh, Proc. Natl.

Acad. Sci. USA 108 (9) (2011) 3809–3814.

.doi.org/10.1016/j.ymeth.2014.02.006

10 A. Huefner et al. / Methods xxx (2014) xxx–xxx

[69] H. Xie, L. Hu, G. Li, Chin. Med. J. (Engl) 123 (8) (2010) 1086–1092.[70] S. Bruno, B. Ardelt, J.S. Skierski, F. Traganos, Z. Darzynkiewicz, Cancer Res. 52

(2) (1992) 470–473.[71] C. Stanwell, A.A. Dlugosz, S.H. Yuspa, Carcinogenesis 17 (6) (1996) 1259–1265.[72] M. Deshmukh, E.M. Johnson, Cell Death Differ. 7 (3) (2000).[73] W. Zhou, M. Choi, D. Margineantu, L. Margaretha, J. Hesson, C. Cavanaugh, C.A.

Blau, M.S. Horwitz, D. Hockenbery, C. Ware, H. Ruohola-Baker, EMBO J. 31 (9)(2012) 2103–2116.

[74] T.E. McGraw, V. Mittal, Nat. Chem. Biol. 6 (3) (2010) 176–177.[75] J. Zhang, E. Nuebel, Cell Stem Cell 11 (5) (2012) 589–595.[76] S. Schlücker, B. Küstner, A. Punge, R. Bonfig, A. Marx, P. Ströbel, J. Raman

Spectrosc. 37 (7) (2006) 719–721.[77] J.G. Huang, T. Leshuk, F.X. Gu, Nano Today 6 (5) (2011) 478–492.

Please cite this article in press as: A. Huefner et al., Methods (2014), http://dx

[78] Y. Yamada, H. Akita, H. Kamiya, K. Kogure, T. Yamamoto, Y. Shinohara, K.Yamashita, H. Kobayashi, H. Kikuchi, H. Harashima, BBA – Biomembranes 1778(2) (2008) 423–432.

[79] B. Kang, M.A. Mackey, M.A. El-Sayed, Causing Cytokinesis Arrest andApoptosis’, J Am Chem Soc 132 (5) (2010) 1517–1519.

[80] P. Nativo, I.A. Prior, M. Brust, ACS Nano 2 (8) (2008) 1639–1644.[81] J.A. Ryan, K.W. Overton, M.E. Speight, C.N. Oldenburg, L. Loo, W. Robarge, S.

Franzen, D.L. Feldheim, Anal. Chem. 79 (23) (2007) 9150–9159.[82] A. Boyd, L. McManus, G. Burke, B. Meenan, J Mater Sci-Mater M 22 (8) (2011)

1923–1930.[83] A.M. Oyarce, P.J. Fleming, Arch. Biochem. Biophys. 290 (2) (1991) 503–510.[84] N.S. Lee, Y.Z. Hsieh, R.F. Paisley, M.D. Morris, Anal. Chem. 60 (5) (1988) 442–

446.

.doi.org/10.1016/j.ymeth.2014.02.006