Parallel Nanometric 3D Tracking of Intracellular Gold Nanorods Using

Jan 29, 2013 - (18) The main obstacle for 3D GNR tracking with TPL is that serial laser scanning and the read-out of a typical confocal microscope ren...
0 downloads 0 Views 3MB Size
Download PDF
Letter pubs.acs.org/NanoLett

Parallel Nanometric 3D Tracking of Intracellular Gold Nanorods Using Multifocal Two-Photon Microscopy Bram van den Broek, Brian Ashcroft, Tjerk H. Oosterkamp, and John van Noort* Leiden Institute of Physics, Leiden University, Niels Bohrweg 2, 2333 CA Leiden, The Netherlands

Downloaded via KAOHSIUNG MEDICAL UNIV on October 30, 2018 at 11:13:55 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

S Supporting Information *

ABSTRACT: We report a novel technique for long-term parallel three dimensional (3D)-tracking of gold nanorods in live cells with nanometer resolution. Gold nanorods feature a strong plasmon-enhanced two-photon luminescence, can be easily functionalized, and have been shown to be nontoxic. These properties make gold nanorods very suitable for in vivo two-photon luminescence microscopy. By rapid multifocal scanning, we combine the advantages of 3D molecular tracking methods using wide-field imaging with the advantages of two-photon microscopy. Isolated gold nanorods can be localized with a resolution of 4 nm in the xy-plane and 8 nm in the z-direction. The polarization-dependence of the two-photon luminescence signal can be used to resolve the angular orientation, even when two gold nanorods are separated by less than the diffraction limit. Individual nanorods in live U2OS cells could be followed in 3 dimensions for over 30 min, with a photon noise limited accuracy, and a time resolution of 50 ms in 2D and 500 ms in 3D. KEYWORDS: Gold nanorods, two-photon imaging, single-particle tracking, nanometry

H

and to minimize cytotoxic effects. Moreover, parallel tracking of multiple closely spaced quantum dots is impeded by their blinking behavior, which dramatically complicates tracing individual particle positions into extended trajectories. Here we use gold nanorods (GNRs) and their plasmonenhanced two-photon luminescence (TPL)9 for in vivo single particle tracking. GNRs as small as 8 nm × 40 nm are detectable with a high signal-to-noise ratio using two-photon microscopy,10 can be easily functionalized,11,12 and have been shown to be nontoxic.13 Functionalization and passivation adds 3−5 nm to the hydrodynamic radius of the GNR.14 The total size is in the range of large protein complexes found in vivo but may still be expected to affect the mobility of conjugated biomolecules, similar to quantum dots. Here we focus on the optical properties of GNRs, which make them very suitable for in vivo TPL confocal microscopy.10 Two-photon luminescence of GNRs has many advantages over other contrast mechanisms like scattering,15 differential interference contrast (DIC),16 and photothermal imaging.17 As TPL scales with the volume of the rod, while scattering intensities scale with the volume squared, the benefits of TPL become increasingly important as particles get smaller. More importantly, for biological studies in scattering media, like in living cells, the absence of background signal in TPL imaging strongly increases the signal-to-noise ratio. Moreover, for tracking applications the better z confinement of two-photon excitation will help to improve 3D localization. Despite these benefits, TLP has to our knowledge only been applied to 2D particle tracking in living cells.18 The main obstacle for 3D

igh-resolution localization microscopy of individual fluorescent molecules has become indispensable for studying biological macromolecules in vitro, as well as inside cells.1 In vivo, a compromise needs to be found between optimizing the optical properties of the labels, to get the best possible positioning accuracy, and reducing the size of the labels, to avoid steric effects that influence the cellular metabolism. Though small organic or biological fluorophores have been used to obtain super-resolution in live cell imaging,2 many practical challenges remain. Time-resolved tracking of the positions of sparse, single fluorophores with subdiffraction resolution can resolve important characteristics of the dynamics of individual proteins in cells but is generally limited to short observation times due to the short period of time before single fluorophores photobleach (a few seconds) and also due to the limited number of photons emitted from these fluorophores, compared to the background autofluorescence.3,4 In contrast, typical cellular activities may take from minutes to hours, and most molecular structures of interest have dimensions in the nanometer range. Moreover, only a small subset of a selected set of biomolecules may participate in the process under study, emphasizing the need for a new imaging technology. To fully exploit the benefits of single-molecule microscopy for resolving molecular dynamics in live cells, it is necessary to develop methods to track multiple probes in parallel, with nanometer resolution and for times ranging from subseconds to hours. Quantum dots have been utilized for prolonged tracking of biomolecules inside cells.5−7 Functionalized quantum dots have a higher brightness and enhanced stability relative to organic dyes or fluorescent proteins. Applications have mostly targeted extracellular membrane proteins,8 avoiding the effects of the large shell that is required to provide biochemical functionality © 2013 American Chemical Society

Received: November 2, 2012 Revised: January 24, 2013 Published: January 29, 2013 980

dx.doi.org/10.1021/nl3040509 | Nano Lett. 2013, 13, 980−986

Nano Letters

Letter

acquisition time for a reasonably homogeneous illumination exceeds 100 ms. Faster scanning is limited by the bandwidth of the mirror, leading to edge artifacts and/or scanner overheating (Supporting Information, Figure S1). Alternatively, we drive the scan mirror with an Archimedean spiral, which significantly reduces the load on the scanner.22 During the spiraling, the scan mirror position and speed are varied such that every spot develops into a broad 2D Gaussian intensity profile:

GNR tracking with TPL is that serial laser scanning and the read-out of a typical confocal microscope renders 3D TPL imaging far too slow (∼minute) as compared to the expected mobility of a single GNR inside a cell. Multifocal Scanning Two-Photon Microscopy. To resolve the bandwidth limitations of two-photon confocal microscopy, we use a multifocal scanning microscope with a CCD based read out and apply this to single GNR tracking in live cells. Figure 1 shows our two-photon multifocal micro-

x = Aτ sin(2πnτ ) with τ =

y = Aτ cos(2πnτ ) ⎛ (t /T )2 − 1 ⎞ ⎟ t /T exp⎜ 2σG 2 ⎠ ⎝

(1)

Here, A is the amplitude of the scanning signal, n the number of spiral arms, and σG the width of the 2D Gaussian profile. The period of this signal is set to the camera exposure time T. The positions of the scanner x and y are sampled with a high rate (typically 100 kHz) to ensure a smooth signal. A flat excitation profile is accomplished when the distance between neighboring excitation spots is less than the the width of the laser focus and when the spiral arm density n is set such that successive spiral arms sufficiently overlap (Supporting Information, Figure S1b). Given the DOE design, the diffraction-limited spots in our setup have a lattice spacing of 2.7 μm, resulting in a total illumination area of approximately 30 × 30 μm2. We characterized the performance of the multifocal microscope and the various scanning patterns by imaging a solution of Rhodamine 6G. Figure 2a shows the hexagonal pattern of spots created by the DOE as well as the resulting illumination profiles. It is clear that spiral scanning yields the most homogeneous illumination, performing increasingly better for lower exposure times, relative to the other methods (Figure 2b). Comparing the histograms of the intensities of the different scanning patterns at 50 ms exposure time (Figure 2c) confirms that spiral scanning results in a very homogeneous wide-field illumination. The benefits of multifocal spiral scanning two-photon microscopy using a DOE are similar to those obtained by using two-dimensional microlens arrays23 but yield a more homogeneous excitation pattern. Multifocal spiral scanning may therefore improve any application requiring rapid two-photon microscopy employing CCD detection. Two-Photon Luminescence of Single Gold Nanorods. High-resolution single-particle tracking in biological applications requires an optimal balance between the brightness of the particle, the background fluorescence, and light-induced damage. The use of near-infrared light for two-photon microscopy has many advantages over one-photon fluorescence: reduced autofluorescence, large penetration depth, and reduced phototoxicity. However, excitation intensity has to be kept low, because photodamage increases nonlinearly for laser powers above a threshold of about 10 mW.24 To characterize the TPL of GNRs in the scanning multifocal microscope we imaged single GNRs immobilized on a glass coverslip. With a typical excitation power of 100 μW per spot (∼5 GW/cm2 peak intensity, 100 kW/cm2 time-averaged) we obtained a constant signal with a signal-to-noise (S/N) ratio of ∼1000 (Supporting Information, Figure S2). For comparison, single Rodamine 6G molecules and single quantum dots yielded a S/N ratio of 7 and 80 at the same excitation power. Furthermore, bleaching of Rhodamine 6G occurred typically after several seconds, whereas the quantum dots displayed characteristic stochastic blinking behavior.

Figure 1. Multifocal two-photon laser scanning microscopy setup. An array of 100 spots is created in the sample plane and the conjugate planes (dashed lines) by passing the beam generated by a Ti:Sapph laser through a diffractive optical element (DOE), in which the zeroorder diffracted beam is blocked. 3D images are created from stacked 2D images recorded by the EMCCD camera. Inset: Representation of a 3D image of a U2OS cell. A detailed description of this setup is provided in the Supporting Information.

scope, in which a 10 × 10 hexagonal array of excitation spots is generated in the image plane by passing the beam through a diffractive optical element (DOE), similar to the setups described in refs 19, 20, and 21. A two-dimensional twophoton fluorescence image is created by scanning the array of spots across the sample plane (x,y) within the CCD exposure time, so that the entire field of view spanned by the array of spots is homogeneously illuminated, comparable to spinning disc microscopy. By scanning multiple slices in the z-direction (see inset of Figure 1), a three-dimensional two-photon image can be recorded with subsecond time resolution. Although much faster than single-beam scanning, standard raster scanning techniques with multiple beams lead to image artifacts such as brick-like patterns at boundaries between scanned subregions, Figure 2a. As a consequence, image quality is compromised. Jureller et al. proposed stochastic scanning; that is, driving the piezo with Gaussian distributed “white noise”.19 Edge effects blend away by overlapping the tails of these Gaussians. However, depending on the subregion size, we observe that hundreds of sampling points are required to attain a reasonably homogeneous excitation profile. Even with high bandwidth scanners (>1000 Hz), the actual minimum 981

dx.doi.org/10.1021/nl3040509 | Nano Lett. 2013, 13, 980−986

Nano Letters

Letter

Figure 2. Comparison of spiral scanning with other scanning methods. (a) Two-photon fluorescence of Rhodamine 6G in solution, imaged 2 μm above the glass−water interface. Three scanning methods were tested for fluorescence homogeneity across the excited field of view. Line profiles along the horizontal lines are indicated below the images. The average fluorescence intensity while scanning is a fifth of the peak intensities without scanning, but as expected the integrated intensity is the same (1790 counts for not scanning vs 1780 counts for spiral scanning). (b) The standard deviations of the intensity distributions calculated from pixel intensities in the central 100 × 100 pixels (18 × 18 μm) region, divided by the mean pixel intensity of this region, for decreasing exposure times, show that the spiral scanning method results in a favorable homogeneity at high frame rates. (c) Measured normalized intensity histograms for spiral, raster, and stochastic scanning at a 25 Hz frame rate confirm that the most homogeneous intensity distribution is achieved with spiral scanning.

To distinguish GNRs that may be separated by less than the diffraction limit, due to the random deposition of GNRs on the glass slide and possible aggregation of the GNRs, we compared TPL images with scanning electron microscopy (SEM) images of the same field of view (Figure 3). Figure 3b shows a zoom-in of a bright TPL spot, selected from the overview TPL image in Figure 3 that, upon SEM inspection, originates from two GNRs located 340 nm apart. Because circularly polarized two-photon excitation could not directly discriminate the two, we captured images with linearly polarized excitation at defined polarization angles to resolve the positions and intensities of the GNRs. Subdiffraction resolution was obtained by fitting two Gaussian peaks, separated by distance R and angle θ. The sum of the intensities of the two TPL peaks accurately follows the expected double cos4-dependence on the excitation polarization angle 10 (Figure 3d). Moreover, the fitted distance and orientation obtained from the TPL image closely match the SEM image. Differences in intensity result from differences in shape of the individual GNRs. The aspect ratio of the used GNRs has a standard deviation of 20%,14 which would translate in absorption maxima between 820 and 1030 nm.25 Variations in the amplitude of the peaks reflect such differences in excitation spectrum and could be used for multicolor imaging. The localization accuracy s that can be achieved by fitting Gaussian intensity profile to a single GNR can be approximated by26

s=

σ2 a2 8πσ 4b2 + + 2 2 N 12N aN

(2)

In which σ corresponds to the standard deviation of the point spread function, N is the total number of collected photons, a the pixel size, and b the background noise. With N being at least 100 times larger for GNRs than for quantum dots or organic dyes, we expect at least 10 times improvement of the localization accuracy. The accuracy can be further boosted by increasing the excitation power, but ultimately it is limited by the absorption and resulting heating of the GNR. As shown in Supporting Information, Figure S3b, the intensity of the TPL signal increases quadratically with excitation power, up to about 50 GW/cm2, above which the GNRs have a tendency to reshape thermally and lose their TPL generated fluorescence for the applied wavelength.27,28 In typical tracking applications we use a factor 100 lower power, which is sufficient for a high S/N ratio. A quantitative estimate of the heating is nontrivial because of the discontinuous excitation by the pulsed laser and the scanning laser foci. By assuming that the heating is proportional to the excitation power and using previous findings that thermal reshaping of GNRs becomes significant at temperatures above 100 °C,27 we estimate that the temperature increase of GNRs at the excitation power used in a typical experiment is only a few degrees. Moreover, calculations have shown that the temperature gradient decays rapidly in a few nanometers.29 Coating of the GNRs with nonabsorbing material, like the PEG/neutravidin coating of the 982

dx.doi.org/10.1021/nl3040509 | Nano Lett. 2013, 13, 980−986

Nano Letters

Letter

Figure 3. (a) Two photon luminescence of gold nanorods immobilized on a glass slide. (b) 500 × 500 nm2 zoom of a single diffraction limited peak. (c) Scanning electron microscopy image of the same field of view as in part b shows two GNRs that are approximately 340 nm apart. (d) The strong polarization dependence of TPL with respect to the orientation of the nanorod allows precise measurement of the position and orientation of each single GNR. Lines represent fits with a cos4 dependence of the intensity of the polarization angle. Blue line shows the cos4 dependence of a single GNR; red line shows the fit to two shifted and rotated cos4 functions, yielding a distance of 346 nm and a difference in orientation of 21°. Experimental details are described in the Supporting Information.

TPL image is fitted with a 3D Gaussian, with independent xyand z widths σ:

GNRs used here, therefore is sufficient to further relieve the heating issue for biological applications. Thus we may expect GNRs to be very compatible with live cell imaging, yielding a significant increase in positional accuracy relative to more traditional single molecule probes. 3D Nanometric Tracking in Live Cells. To do a real test of the in vivo accuracy of single GNR tracking with TPL, it is necessary to measure inside cells. This affects the localization accuracy by an increased background signal, originating from the autofluorescence of the cell and the out of focus fluorescence. Furthermore, the dynamics of the GNR will put limits on the observation time, and perhaps most importantly, GNRs can move in 3 dimensions and must therefore be localized in 3D. Two-photon excitation contributes to a better accuracy by the reduced cellular autofluorescence and by the improved excitation confinement in the z-direction. To obtain three-dimensional information we collect stacks of images using a piezo-driven objective. 2D slices separated by 400 nm are imaged at 25 frames per second. In this way, a 30 × 30 × 10 μm3 volume per second is covered, sufficient to image an entire mammalian cell adhered to a glass surface. Figure 4a shows a 2D projection of a stack of TPL slices of GNRs incorporated by endocytosis into a live U2OS cell, overlaid with a transmission image of the cell. Each peak in the

⎛ (x − x )2 + (y − y )2 ⎞ 0 0 ⎟ f (x , y , z) = A exp⎜⎜ − ⎟ 2 2 σ ⎝ ⎠ xy ⎛ (z − z )2 ⎞ 0 ⎟ exp⎜ − 2σz 2 ⎠ ⎝

(3)

In the Supporting Information, Figure S3a, a one-dimensional representation of the 3D luminescence signal of a single GNR is plotted (7 × 7 × 7 pix), along with the (six parameter) fit and the residual. We determined the average localization accuracy s of GNRs to be better than 5 nm for x0 and y0 and 10 nm for z0 (Figure 4b). The accuracy was obtained in two independent ways. First, we calculated the average standard error in the fit positions of more than 50 GNRs inside or adhered to a living U2OS cell, for 300 3D frames. Second, we computed the standard deviations of the obtained positions of GNR immobilized in a PAGE gel, corrected for drift, in a timelapse experiment. Both methods resulted in similar estimates of the accuracy, which is a several-fold enhancement over other recent 3D nanoparticle tracking methods.30−32 The enhanced accuracy of single particle tracking with GNRs can be attributed 983

dx.doi.org/10.1021/nl3040509 | Nano Lett. 2013, 13, 980−986

Nano Letters

Letter

reproducibility of the localization of the GNRs. Perhaps equally important for recording long time traces is the absence of blinking and bleaching. This makes connecting successive coordinates straightforward. We have recorded continuous trajectories of GNRs in live cells for more than 30 min. During this time we observed no morphological changes of the cell that indicate deleterious effects of continuous illumination, heating, or cytotoxicity. Figure 5a−c show three different trajectories of single PEG-coated GNRs moving in the same cell (see Supporting Movies 1 and 2). Although many rods seem at first sight stationary in or on the cell, the high-resolution 3D trajectories show that these GNR undergo diverse nanoscale movement. A mean-squared displacement (MSD) analysis of the measured trajectories, shown in Figure 5d, reveals three types of motion: immobile, Brownian diffusion, and diffusion combined with active transport. Some GNRs show fast directional movement over distances spanning the entire cell (Sup. Figure S4a and b), with velocities up to several μm/s. Other GNRs show pure diffusional motion in three dimensions, with a broad distribution of diffusion constants, varying between 10 and 104 nm2/s. Though the MSD increases linearly, indicating free diffusion, the diffusion constant is about 103−106 times smaller than the expected diffusion constant for GNR in an aqueous solution. The observed reduced diffusion coefficient is consistent with the uptake of GNRs in larger vesicles,33 which limits the mobility of the GNRs. In some occasions these 3D trajectories are also interspersed with clear periods of transport (Sup. Figure S4c and d). In such instances it is likely that the GNR inside a vesicle is transported by molecular motors which move the vesicle along the microtubule network, similar as observed for quantum dot clusters.34 In conclusion, long-term trajectory analysis of single GNRs can yield a nanometric description of diverse cellular activities. The MSD analysis also gives an independent check of the positional accuracy. For free diffusion in n dimensions the MSD can be described as:

Figure 4. (a) The summed intensity of a TPL image stack of GNRs (blue) in live U2OS cells superimposed on a transmission image (gray scale). Several tens of GNRs are incorporated in the cell by endocytosis. (b) The standard errors of the fitted positions in timelapse movies of hundreds of GNRs measured in live cells show an accuracy better than 10 nm in three dimensions. (c) 3D reconstruction of the trajectories of 30 GNRs in a live HeLa cell that were tracked simultaneously. Green clusters of dots, highlighted by the green circles, represent GNRs that display little movement and were confined to the z = 0 plane, probably due to sticking to the glass slide. Red and cyan clusters of points show examples of intracellular trajectories.

to the high brightness of the TLP signal, as described in eq 2. Because the surface plasmon resonance peak of GNR redshifts with increasing aspect ratio of the rod, the number of detected photons at a certain excitation wavelength varies with shape of each GNR. The narrow excitation spectrum of the GNRs in combination with a spread in the aspect ratio of the GNRs results in a variation in the intensity between different GNRs. This explains the wide tail in the accuracy distribution in Figure 4c. In the Supporting Information, Figure S3c, we compare the theoretically expected localization accuracy with the experimental accuracy obtained from 3D fits of a large number of GNRs with different intensities inside a cell. Indeed the accuracy decreases with the square root of the brightness, following eq 2, but is slightly offset by several nm compared to the theoretical limit. Fitting the peaks in single slices yields the 2D position, which closely follows eq 2. The offset in the 3D accuracy may be caused by diffusion of the GNRs during acquisition of the z-stack or by deviations of the point spread function from a 3D Gaussian. Nevertheless, with the TPL imaging of GNRs it is possible to track nanometer-sized objects in live cells with nanometer resolution. Time Analysis of Intracellular GNR Mobility. The exceptional localization accuracy in combination with the fast two-photon imaging allows us to track the 3D movement of GNR inside the cell with exquisite detail. Figure 4c shows a 3D reconstruction of the positions of 30 GNRs that were acquired in parallel. Note that several GNRs appear to be stuck to the glass slide (plotted in green), highlighting the excellent

MSD(t ) = 2nDt + v 2t 2 +

∑ 2sn2 n

(4)

The last term in eq 4 represents the inaccuracy of the localization. Stochastic errors in the position add a constant factor to the MSD. From the intercept at t = 0 we can independently determine the combined accuracy sn in x, y, and z. For the trajectories displayed in Figure 5a−c s3 is 7−10 nm; see the inset of Figure 5d. GNR luminescence is highly dependent on the orientation with respect to the polarization of the excitation light, as shown in Figure 3. Combined with the 3D tracking this feature offers the appealing possibility of sensing the rotational freedom of the rods in the cell. As a proof of principle we monitored the TPL intensity fluctuations of intracellular GNRs in a single zposition at 25 Hz, with linearly as well as circularly polarized excitation light (Figure 5e). Indeed linearly polarized excitation yields highly fluctuating peak intensities for some GNRs, whereas circularly polarized excitation displays a constant intensity. Free GNRs in aqueous solution have a rotational diffusion constant on the order of 104 s−1, which is too fast to resolve at 25 Hz. The observed fluctuations therefore indicate that the GNRs are attached to cellular components, consistent with the reduced lateral diffusion constant that we observed. The strong polarization dependence allows for detailed in vivo analysis of orientational changes of GNRs attached to 984

dx.doi.org/10.1021/nl3040509 | Nano Lett. 2013, 13, 980−986

Nano Letters

Letter

Figure 5. Typical 30 min trajectories of single GNRs in a live HeLa cell, display (a) immobilized GNRs, (b) difussive motion, and (c) active transport. (d) Mean-squared displacement analysis of the trajectories shown in a−c and fits to eq 4. The inset shows the offset at t = 0, which represents the accuracy of the localization. (e) Two representative intensity traces of a single GNR in a live U2OS cell, measured with circularly polarized light (blue) and linearly polarized light (red). Large fluctuations when using linearly polarized light occur due to rotational diffusion.

trapping beam and/or may be trapped together. Accurate and prolonged tracking of the GNRs in such a complex environment is a first step toward the use of optical tweezers in vivo. The GNRs studied here were incorporated in the cells by endocytosis, which encapsulates the GNRs in larger vesicles and thus reduces their dynamics. Other techniques may be used to incorporate GNRs in the cell,38 such as microinjection, which may allow GNRs, having a size similar to that of large protein complexes, to freely diffuse inside the cell. The time resolution of our technique would not be sufficient to follow such highly mobile GNRs. This may be used as an advantage for long-term tracking applications, allowing a clear distinction of GNRs specifically conjugated to larger structures from unbound GNRs. Alternatively, faster 3D tracking methods based on astigmatism30 or bifocal imaging39 can be easily incorporated in the current setup, relieving the need to acquire multiple stacks of images. Astigmatism or bifocal imaging generally requires illumination of a wide z-section, which is not compatible with two-photon excitation. We note though that the accuracy of TPL GNR tracking is 2 orders of magnitude better than the illumination depth, allowing for a compromise between dynamic range and tracking speed. When a single plane is imaged, fitting of a 3D Gaussian as described here will not yield a z-coordinate, but astigmatism or bifocal imaging within this plane should still work. Therefore, one can achieve a much better time resolution at the expense of z-range, which may be useful in certain applications. We also note that alternative two-photon excitation strategies, using acousto-optic deflectors40 or spatial light modulators,41 MEMS-based scanners,42 or fast line- or plane-

biomolecules with rotational constraint that are anticipated to show such behavior such as kinesins or other molecular motors. In conclusion, we have developed a multifocus two-photon microscope with which many individual GNRs can be tracked in 3D over extended time periods and with nanometer accuracy. These developments represent a 2 orders of magnitude increase in intensity and 1 order of magnitude increase in localization accuracy compared to that of quantum dots under the same conditions. We have shown that TPL imaging of single GNRs is compatible with live cell imaging. The low cytotoxicity and the versatile surface chemistry of gold allow straightforward conjugation of biomolecules. Though the TPL brightness of the GNRs allows for very small illumination powers, much smaller than previously established damage thresholds, we cannot exclude light-induced physiological changes. Morphological changes are largely absent in our measurements; see for example Supporting Movie 3. A more thorough analysis of such effects however requires further study and will be subject of future investigations. Interestingly, GNRs are amendable to optical tweezers manipulation,35,36 making GNRs a promising tool for nanobiology inside the living cell. One of the obstacles for using GNRs in optical tweezers is the heating of a trapped GNR, which was measured to be 70 K under typical off-resonance trapping conditions,36 but can be up to several 100 K at resonance.37 We note that we use 3 orders of magnitude less power per spot for TPL imaging and that spiral scanning further reduces the time-averaged power, leading to a strongly reduced heating compared to trapping conditions. Another experimental challenge for trapping GNRs is the presence of various cellular structures in the cell that can distort the 985

dx.doi.org/10.1021/nl3040509 | Nano Lett. 2013, 13, 980−986

Nano Letters

Letter

illumination methods43 may be used in combination with 3D GNR imaging to speed up multiparticle tracking. Here we have shown that multifocal two-photon spiral scanning yields shot noise-limited localization accuracy inside cells due to the low background signal, making TPL imaging of GNRs an ideal technique for high-resolution 3D particle tracking in living cells.

■

(19) Jureller, J. E.; Kim, H. Y.; Scherer, N. F. Opt. Express 2006, 14, 12482−12483. (20) Zhang, R.; Rothenberg, E.; Fruhwirth, G.; Simonson, P. D.; Ye, F.; Golding, I.; Ng, T.; Lopes, W.; Selvin, P. R. Nano Lett. 2011, 11, 4074−8. (21) Sacconi, L.; Froner, E.; Antolini, R.; Taghizadeh, M. R.; Choudhury, A.; Pavone, F. S. Optics Letters 2003, 28, 1918−20. (22) Helmchen, F.; Go, W. Nat. Methods 2007, 4, 73−79. (23) Squier, J.; Brakenhoff, G. J.; Buist, A. H.; Mu, M. J. Microsc. 1998, 192, 217−226. (24) Hopt, A.; Neher, E. Biophys. J. 2001, 80, 2029−36. (25) Link, S.; Mohamed, M. B.; El-Sayed, M. A. J. Phys. Chem. B 1999, 103, 3073−3077. (26) Thompson, R. E.; Larson, D. R.; Webb, W. W. Biophys. J. 2002, 82, 2775−83. (27) Mohamed, M. B.; Ismail, K. Z.; Link, S.; El-Sayed, M. A. J. Phys. Chem. B 1998, 102, 9370−9374. (28) Liu, Y.; Mills, E. N.; Composto, R. J. J. Mater. Chem. 2009, 19, 2704. (29) Ekici, O.; Harrison, R. K.; Durr, N. J.; Eversole, D. S.; Lee, M.; Ben-Yakar, A. J. Phys. D: Appl. Phys. 2008, 41, 185501. (30) Holtzer, L.; Meckel, T.; Schmidt, T. Appl. Phys. Lett. 2007, 90, 053902. (31) Ram, S.; Prabhat, P.; Chao, J.; Ward, E. S.; Ober, R. J. Biophys. J. 2008, 95, 6025−43. (32) Levi, V.; Ruan, Q.; Gratton, E. Biophys. J. 2005, 88, 2919−28. (33) Chithrani, B. D.; Chan, W. C. W. Nano Lett. 2007, 7, 1542−50. (34) Nan, X.; Sims, P. A.; Chen, P.; Xie, X. S. J. Phys. Chem. B 2005, 109, 24220−4. (35) Selhuber-Unkel, C.; Zins, I.; Schubert, O.; Sönnichsen, C.; Oddershede, L. B. Nano Lett. 2008, 8, 2998−3003. (36) Ruijgrok, P.; Verhart, N.; Zijlstra, P.; Tchebotareva, a.; Orrit, M. Phys. Rev. Lett. 2011, 107, 1−4. (37) Ma, H.; Bendix, P. M.; Oddershede, L. B. Nano Lett. 2012, 12, 3954−60. (38) Delehanty, J. B.; Mattoussi, H.; Medintz, I. L. Anal. Bioanal. Chem. 2009, 393, 1091−1105. (39) Toprak, E.; Balci, H.; Blehm, B. H.; Selvin, P. R. Nano Lett. 2007, 7, 2043−5. (40) Iyer, V.; Hoogland, T. M.; Saggau, P. J. Neurophysiol. 2006, 95, 535−45. (41) Nikolenko, V.; Watson, B. O.; Araya, R.; Woodruff, A.; Peterka, D. S.; Yuste, R. Front. Neural Circuits 2008, 2, 5. (42) Piyawattanametha, W.; Barretto, R. P. J.; Ko, T. H.; Flusberg, B. A.; Cocker, E. D.; Ra, H.; Lee, D.; Solgaard, O.; Schnitzer, M. J. Opt. Lett. 2006, 31, 2018−20. (43) Vaziri, A.; Tang, J.; Shroff, H.; Shank, C. V. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 20221−6.

ASSOCIATED CONTENT

* Supporting Information S

Methods and figures (pdf); a 30 × 30 μm stack of frames showing single PEG-coated CNRs moving in a live U2OS cell (movie 1, avi); a 30 × 30 μm time trace of single PEG-coated CNRs moving in a live U2OS cell (movie 2, avi); 30 × 30 μm transmission image of a live U2OS cell (movie 3, avi). This material is available free of charge via the Internet at http:// pubs.acs.org.

■ ■

AUTHOR INFORMATION

Notes

The authors declare no competing financial interest.

ACKNOWLEDGMENTS This work was supported by a Netherlands SmartMix grant and the partner organizations of NIMIC (http://www.nimicproject.com), a public-private program, and by the Foundation for Fundamental Research on Matter (FOM), which is part of the Dutch Organization for Scientific Research (NWO). We thank Bas van Vliet, Peter Zijlstra, and Michel Orrit for helpful discussions.

■

REFERENCES

(1) Moerner, W. E. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 12596− 602. (2) Schmidt, T.; Schütz, G. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 2926−2929. (3) Shaner, N. C.; Steinbach, P. A.; Tsien, R. Y. Nat. Mater. 2005, 2, 905−909. (4) Lommerse, P. H. M.; Snaar-Jagalska, B. E.; Spaink, H. P.; Schmidt, T. J. Cell Sci. 2005, 118, 1799−809. (5) Michalet, X.; Pinaud, F. F.; Bentolila, L. A.; Tsay, J. M.; Doose, S.; Li, J. J.; Sundaresan, G.; Wu, A. M.; Gambhir, S. S.; Weiss, S. Science (New York, N.Y.) 2005, 307, 538−44. (6) Pinaud, F.; Clarke, S.; Sittner, A.; Dahan, M. Nat. Methods 2010, 7, 275−85. (7) Medintz, I. L.; Uyeda, H. T.; Goldman, E. R.; Mattoussi, H. Nat. Mater. 2005, 4, 435−46. (8) Pinaud, F.; Michalet, X.; Iyer, G.; Margeat, E.; Moore, H.-P.; Weiss, S. Traffic (Copenhagen, Denmark) 2009, 10, 691−712. (9) Wang, D.-S.; Hsu, F.-Y.; Lin, C.-W. Opt. Express 2009, 17, 11350−9. (10) Wang, H.; Huff, T. B.; Zweifel, D. A.; He, W.; Low, P. S.; Wei, A.; Cheng, J.-X. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 15752−6. (11) Durr, N. J.; Larson, T.; Smith, D. K.; Korgel, B. A.; Sokolov, K.; Ben-Yakar, A. Nano Lett. 2007, 7, 941−5. (12) Nanorods, A. G.; Pissuwan, D.; Valenzuela, S. M.; Miller, C. M.; Cortie, M. B. Nano Lett. 2007, 7, 3808−3812. (13) Lewinski, N.; Colvin, V.; Drezek, R. Small 2008, 4, 26−49. (14) Personal communication; Nanopartz Inc.: Loveland, CO, 2012. (15) Anderson, L. J. E.; Mayer, K. M.; Fraleigh, R. D.; Yang, Y.; Lee, S.; Hafner, J. H. J. Phys. Chem. C 2010, 114, 11127−11132. (16) Wang, G.; Sun, W.; Luo, Y.; Fang, N. J. Am. Chem. Soc. 2010, 132, 16417−22. (17) Boyer, D.; Tamarat, P.; Maali, A.; Lounis, B.; Orrit, M. Science 2002, 297, 1160−3. (18) Huff, T. B.; Hansesn, M. N.; Zhao, Y.; Cheng, J.-X.; Wei, A. Langmuir 2007, 23, 1596−9. 986

dx.doi.org/10.1021/nl3040509 | Nano Lett. 2013, 13, 980−986